Ghosh S. Mathematics and Computer Science Vol 2. 2023

Mathematics and Computer
Science Volume 2
Scrivener Publishing
100 Cummings Center, Suite 541J
Beverly, MA 01915-6106
Advances in Data Engineering and Machine Learning
Series Editors: Niranjanamurthy M, PhD, Juanying XIE, PhD, and Ramiz Aliguliyev, PhD
Scope: Data engineering is the aspect of data science that focuses on practical applications of data collection and
analysis. For all the work that data scientists do to answer questions using large sets of information, there
have to be mechanisms for collecting and validating that information. Data engineers are responsible for
finding trends in data sets and developing algorithms to help make raw data more useful to the enterprise.
It is important to have business goals in line when working with data, especially for companies that handle
large and complex datasets and databases. Data Engineering Contains DevOps, Data Science, and Machine
Learning Engineering. DevOps (development and operations) is an enterprise software development phrase
used to mean a type of agile relationship between development and IT operations. The goal of DevOps is
to change and improve the relationship by advocating better communication and collaboration between
these two business units. Data science is the study of data. It involves developing methods of recording,
storing, and analyzing data to effectively extract useful information. The goal of data science is to gain
insights and knowledge from any type of data — both structured and unstructured.
Machine learning engineers are sophisticated programmers who develop machines and systems that can
learn and apply knowledge without specific direction. Machine learning engineering is the process of
using software engineering principles, and analytical and data science knowledge, and combining both
of those in order to take an ML model that’s created and making it available for use by the product or
the consumers. “Advances in Data Engineering and Machine Learning Engineering” will reach a wide
audience including data scientists, engineers, industry, researchers and students working in the field of
Data Engineering and Machine Learning Engineering.
Publishers at Scrivener
Martin Scrivener (martin@scrivenerpublishing.com)
Phillip Carmical (pcarmical@scrivenerpublishing.com)
Mathematics and Computer
Science Volume 2
Edited by
Sharmistha Ghosh
M. Niranjanamurthy
Krishanu Deyasi
Biswadip Basu Mallik
and
Santanu Das
This edition first published 2023 by John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, USA
and Scrivener Publishing LLC, 100 Cummings Center, Suite 541J, Beverly, MA 01915, USA
© 2023 Scrivener Publishing LLC
For more information about Scrivener publications please visit www.scrivenerpublishing.com.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or
transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or other-
wise, except as permitted by law. Advice on how to obtain permission to reuse material from this title
is available at http://www.wiley.com/go/permissions.
Wiley Global Headquarters

111 River Street, Hoboken, NJ 07030, USA
For details of our global editorial offices, customer services, and more information about Wiley prod-
ucts visit us at www.wiley.com.
Limit of Liability/Disclaimer of Warranty

While the publisher and authors have used their best efforts in preparing this work, they make no rep
resentations or warranties with respect to the accuracy or completeness of the contents of this work and
specifically disclaim all warranties, including without limitation any implied warranties of merchant-
ability or fitness for a particular purpose. No warranty may be created or extended by sales representa
tives, written sales materials, or promotional statements for this work. The fact that an organization,
website, or product is referred to in this work as a citation and/or potential source of further informa
tion does not mean that the publisher and authors endorse the information or services the organiza
tion, website, or product may provide or recommendations it may make. This work is sold with the
understanding that the publisher is not engaged in rendering professional services. The advice and
strategies contained herein may not be suitable for your situation. You should consult with a specialist
where appropriate. Neither the publisher nor authors shall be liable for any loss of profit or any other
commercial damages, including but not limited to special, incidental, consequential, or other damages.
Further, readers should be aware that websites listed in this work may have changed or disappeared
between when this work was written and when it is read.
Library of Congress Cataloging-in-Publication Data
ISBN 978-1-119-89632-6
Front cover images supplied by Wikimedia Commons

Cover design by Russell Richardson
Set in size of 11pt and Minion Pro by Manila Typesetting Company, Makati, Philippines
Printed in the USA
10 9 8 7 6 5 4 3 2 1
Contents
Preface xvii
1 A Comprehensive Review on Text Classification and
Text Mining Techniques Using Spam Dataset Detection 1
Tamannas Siddiqui and Abdullah Yahya Abdullah Amer
1.1 Introduction 2
1.2 Text Mining Techniques 3
1.2.1 Data Mining 3
1.2.2 Information Retrieval 4
1.2.3 Natural Language Processing (NLP) 5
1.2.4 Information Extraction 5
1.2.5 Text Summarization 6
1.2.6 Text Categorization 7
1.2.7 Clustering 7
1.2.8 Information Visualization 7
1.2.9 Question Answer 8
1.3 Dataset and Preprocessing Steps 9
1.3.1 Text Preprocess 9
1.4 Feature Extraction 9
1.4.1 Term Frequency – Inverse Document Frequency 10
1.4.2 Bag of Words (BoW) 10
1.5 Supervised Machine Learning Classification 11
1.6 Evaluation 11
1.7 Experimentation and Discussion Results for Spam
Detection Data 11
1.8 Text Mining Applications 13
1.9 Text Classification Support 13
1.9.1 Health 13
1.9.2 Business and Marketing 14
1.9.3 Law 14
v
vi Contents
1.10 Conclusions 14
References 15
2 Study of Lidar Signals of the Atmospheric Boundary Layer
Using Statistical Technique 19
Kamana Mishra and Bhavani Kumar Yellapragada
2.1 Introduction 19
2.2 Methodology 21
2.2.1 A Statistical Approach to Determine the CBLH 21
2.2.2 A Statistical Approach to Determine the Best
Fit Distribution to the Backscatter Signals of the
Lidar Dataset 22
2.3 Mathematical Background of Method 23
2.4 Example and Result 24
2.5 Conclusion and Future Scope 27
Acknowledgement 28
References 28
Annexure 30
3 Optimal Personalized Therapies in Colon Cancer Induced
Immune Response using a Fokker-Planck Framework 33
Souvik Roy and Suvra Pal
3.1 Introduction 33
3.2 The Control Framework Based on Fokker-Planck Equations 35
3.3 Theoretical Results 39
3.4 Numerical Schemes 41
3.5 Results 43
3.6 Conclusion 44
Acknowledgments 45
References 45
4 Detection and Classification of Leaf Blast Disease using
Decision Tree Algorithm in Rice Crop 49
Sarvesh Vishwakarma and Bhavna Chilwal
4.1 Introduction 49
4.2 Proposed Methodology 51
4.3 Result Analysis 52
4.4 Conclusion 55
4.5 Future Work 55
References 56
Contents vii
5 Novel Hybrid Optimal Deep Network and Optimization

Approach for Human Face Emotion Recognition 59
J. Seetha, M. Ayyadurai and M. Mary Victoria Florence
5.1 Introduction 60
5.2 Related Work 61
5.3 System Model and Problem Statement 62
5.4 Proposed Model 63
5.4.1 Preprocessing Stage 64
5.4.2 Knowledge Based Face Detection 64
5.4.3 Image Resizing 64
5.4.4 Feature Extraction 64
5.5 Proposed HDC-GEN Classification 65
5.6 Result and Discussion 68
5.6.1 Performance Metrics Evaluation 69
5.6.2 Comparative Analysis 70
5.7 Conclusion 74
References 74
6 An Application of Information Technology in
Adaptive Leadership of Ministry of Ayush During Pandemic
of Covid 19: A Case Study 77
Vikram Singh, Shikha Kapoor and Sandeep Kumar Gupta
6.1 Introduction 77
6.2 Ministry of AYUSH 78
6.3 Leadership Principles and Practices by Ministry
of AYUSH During Covid-19 79
6.4 Effective Communication 79
6.5 Sharing of Resources 80
6.6 Shared Decision Making 81
6.7 Training of Manpower 81
6.8 Use of IT Platform 81
6.9 Finding Opportunities for R&D During the Crisis 83
6.10 Collaborating with Stakeholders for International Day
of Yoga (IDY) 84
6.11 Providing Hope When Nothing Seemed
to be Working 87
6.12 Leveraging Old Knowledge 87
6.13 Conclusion 88
References 88
viii Contents
7 Encoder-Decoder Models for Protein Secondary Structure

Prediction 91
Ashish Kumar Sharma and Rajeev Srivastava
7.1 Introduction 91
7.2 Literature Review 93
7.3 Experimental Work 93
7.3.1 Data Set 93
7.3.2 Proposed Methodology 94
7.3.3 Data Preprocessing 94
7.3.4 Long Short Term Memory 94
7.4 Results and Discussion 96
7.5 Conclusion 97
References 98
8 Hesitancy, Awareness, and Vaccination: A Computational
Analysis on Complex Networks 101
Dibyajyoti Mallick, Aniruddha Ray, Ankita Das
and Sayantari Ghosh
8.1 Introduction 101
8.2 Model Formulation 103
8.3 Model Analysis on Complex Network 106
8.3.1 Effect of Vaccination Rate 107
8.3.2 Effect of Negative Rumors 108
8.3.3 Effect of Positive Peer Influence 108
8.4 Conclusions and Perspectives 111
References 112
9 Propagation of Seismic Waves in Porous Thermoelastic
Semi-Infinite Space with Impedance Boundary Conditions 115
Annu Rani and Dinesh Kumar Madan
9.2 Basic Equations 116
9.3 Problem Formulation 118
9.4 Reflection at the Free Surface 121
9.4.1 Boundary Conditions 124
9.4.2 Energy Ratios 126
9.5 Numerical Results and Discussion 127
9.6 Conclusion 133
References 134
Contents ix
10 IoT Based Ensemble Predictive Techniques to Determine

the Student Observing Analysis through E-Learning 137
Rufia Thaseen I., S. Shahar Banu and Sudha Rajesh
10.2 Review of Literature 140
10.2.1 Objectives of the Study 142
10.3 Methodology 142
10.4 Analysis and Interpretation 143
10.5 Findings and Conclusion 147
References 148
11 Modelling and Analysis of a Congestion Dependent Queue
with Bernoulli Scheduled Vacation Interruption and
Client Impatience 151
K. Jyothsna and P. Vijaya Kumar
11.2 Model Overview 154
11.3 Model Analysis 155
11.3.1 Pre-Arrival Epoch Probabilities 157
11.3.2 Arbitrary Epoch Probabilities 162
11.4 Special Cases 163
11.5 Performance Metrics 163
11.6 Numerical Outcomes 164
11.7 Conclusion 169
References 169
12 Resource Allocation Determines Alternate Cell Fate in
Bistable Genetic Switch 173
Priya Chakraborty and Sayantari Ghosh
12.2 Model Formulation 176
12.3 Result Section 178
12.3.1 Pitchfork Bifurcation in Genetic Toggle 178
12.3.1.1 Resource Affinity Regulates the
Symmetry of Pitchfork Bifurcation 178
12.3.1.2 Availability of Total mRNA Pool
Regulates the Symmetry of Pitchfork
Bifurcation 180
12.3.1.3 Total Resource Availability Regulates
the Point of Bifurcation in the System 180
x Contents
12.3.2 Saddle Node Bifurcation in Genetic Toggle 180

12.3.2.1 Resource Distribution Regulates the
Point of Bifurcation in Toggle Switch 182
12.3.2.2 Region of Interest in Toggle Switch is
Significantly Regulated by Resource
Allocation 182
12.3.2.3 Total Resource Availability T Regulates
Saddle Node Bifurcation Curve 183
12.4 Conclusion 183
Acknowledgement 184
References 184
13 A Hybrid Approach to Ontology Evaluation 187
Aastha Mishra and Preetvanti Singh
13.2 Background 188
13.3 The Developed OntoEva Method 189
13.4 Ontology Selection for Epilepsy Disorder 190
13.4.1 Accuracy 191
13.4.2 Adaptability 191
13.4.3 Clarity 191
13.4.4 Completeness 192
13.4.5 Conciseness 192
13.4.6 Consistency 192
13.4.7 Organizational Fitness 192
13.5 Results 201
13.6 Comparison of Ontologies 201
13.7 Conclusion 202
References 203
14 Smart Health Care Waste Segregation and Safe Disposal 205
R.M. Bommi, Sami Venkata Sai Rajeev, Sarvepalli Navya,
Veluru Sai Teja and Uppala Supriya
14.2 Related Works 207
14.3 System Architecture 210
14.3.1 Wrapping 212
14.3.2 Incinerator 213
14.3.3 Conveyor Cleaning System 213
14.3.4 Circuit Diagram 213
14.3.5 Optimal Path Planning Algorithm for Waste
Collection 214
Contents xi
14.4 Methodology 214

14.5 Mobile App 217
14.6 Conclusions and Future Works 218
Declarations 219
Availability of Data and Materials 219
Competing Interests 219
Author’s Contribution 219
References 219
15 Investigation of Viscoelastic Magnetohydrodynamics (MHD)
Flow Over an Expanded Lamina Surrounded in a Permeable
Media 223
Hiranmoy Mondal, Arindam Sarkar and Raj Nandkeolyar
15.1.1 Literature Review 223
15.1.2 Nomenclature 224
15.2 Formulation of the Problem 226
15.2.1 Analytical Solution 227
15.2.2 Numerical Methods (Spectral Quasi-Linearization
Methods) 228
15.3 Result and Argument 230
15.4 Conclusion 235
References 236
16 Quickest Multi-Commodity Contraflow with Non-Symmetric
Traversal Times 239
Shiva Prakash Gupta, Urmila Pyakurel
and Tanka Nath Dhamala
16.2 Preliminaries with Flow Models 242
16.2.1 Mathematical Model with Contraflow 242
16.3 QMCCF with Non-Symmetric Transit Times 244
16.3.1 Approximation Approach for the QMCCF 245
16.4 Conclusions 248
Acknowledgments 248
References 248
17 A Mathematical Representation for Deteriorating Goods with
a Trapezoidal-Type Demand, Shortages and Time Dependent
Holding Cost 251
Ruma Roy Chowdhury
xii Contents
17.2 Assumptions and Notations 253

17.2.1 Assumptions 253
17.2.2 Notation 254
17.2.3 Decision Variables 255
17.3 Formulation and Solution 255
17.3.1 Case 1 i) 0 < ν < t1 < t2 < t3 < T 255
17.3.2 Case 2 ii) 0 < ta < ν < tb < tc < T 257
17.3.3 Case 3 iii) 0 < ta < tb < ν < tc < T 259
17.4 Numerical Example 261
17.5 Discussion 261
17.6 Inference 262
References 262
18 An Amended Moth Flame Optimization Algorithm Based on
Fibonacci Search Approach for Solving Engineering Design
Problems 265
Saroj Kumar Sahoo and Apu Kumar Saha
18.2 Classical MFO Algorithm 268
18.3 Proposed Method 270
18.4 Results and Discussions on IEEE CEC 2019
Benchmark Problems 273
18.5 Real-Life Applications 277
18.5.1 Optimal Gas Production Capacity Problem 277
18.5.2 Three-Bar Truss Design (TSD) Problem 277
18.6 Conclusion with Future Studies 278
References 279
19 Image Segmentation of Neuronal Cell with Ensemble
Unet Architecture 283
Kirtan Kanani, Aditya K. Gupta, Ankit Kumar Nikum,
Prashant Gupta and Dharmik Raval
19.2 Methods 285
19.3 Dataset 285
19.4 Implementation Details 286
19.5 Evaluation Metrics 286
19.6 Result 286
19.7 Conclusion 289
References 289
Contents xiii
20 Automorphisms of Some Non-Abelian p−Groups of Order p4 291

Muniya Sansanwal, Harsha Arora and Mahender Singh
20.2 Categorization of p-Groups with Order p4 292
20.3 Number of Automorphisms of Some
Non-Abelian Groups of Order p4 293
20.3.1 Computation of Automorphisms of Group G9 293
20.3.2 Computation of Automorphisms of Group G10 297
20.3.3 Computation of Automorphisms of G11 300
References 304
21 Viscoelastic Equation of p-Laplacian Hyperbolic Type
with Logarithmic Source Term 305
Nazlı Irkıl and Erhan Pişkin
21.2 Preliminaries 307
21.3 Global Existence Result 314
21.4 Blow Up Results of the Solution for Equation (21.1) 317
References 324
22 Flow Dynamics in Continuous-Time with Average Arc Capacities 327
Badri Prasad Pangeni and Tanka Nath Dhamala
22.2 Literature Review 329
22.3 Failure in Extension of AP to AAP 329
22.4 Formulation 331
22.5 Conclusion 334
Acknowledgment 334
References 334
23 Analysis of a Multiserver System of Queue-Dependent
Channel Using Genetic Algorithm 337
Anupama and Chandan Kumar
23.2 Description of the Model 338
23.3 Notations 338
23.4 Steady State Equations 340
23.4.1 Performance Characteristics of the System 342
23.4.2 Queue Length Evaluations at Different Epochs 343
23.4.3 Leisure Period and Working Period Length 343
xiv Contents
23.4.4 Cost of the System 343

23.5 Conclusions 347
References 347
24 An Approach to Ranking of Single Valued Neutrosophic
Fuzzy Numbers Based on (α, β, γ) Cut Sets 349
Sunayana Saikia
24.2 Definition and Representations 351
24.3 Proposed Method 355
24.4 Theorems 356
24.5 Numerical Examples 357
24.6 Conclusion 358
References 359
25 Performance Analysis of Database Models Based on Fuzzy
and Vague Sets for Uncertain Query Processing 363
Sharmistha Ghosh and Surath Roy
25.2 Basic Definitions 364
25.2.1 Fuzzy Set 365
25.2.2 Vague Set 365
25.2.3 Similarity Measure 365
25.3 Algorithm to Generate Membership Values 366
25.4 Real Life Applications 368
25.5 Conclusion 382
References 382
26 Estimating Error of Signals by Product Means
( N , pn , qn ) (C, 2) of the Fourier Series in a
r
W(L , ξ (t))(r ≥ 1) Class 385
Pankaj Tiwari and Aradhana Dutt Jauhari
26.1.1 Definition 1 386
26.1.5 Remark 1 387
26.2 Known Result 388
26.2.1 Theorem 1 388
26.3 Main Theorem 389
Contents xv
26.3.1 Theorem 2 389

26.4 Some Auxiliary Results 390
26.4.1 Lemma-1 390
26.4.2 Lemma-2 390
26.5 Theorem’s Proof 391
26.6 Applications 394
26.6.1 Cor. 1 394
26.6.2 Cor. 2 394
26.7 Conclusion 394
Acknowledgement 395
References 395
About the Editors 397
Index 401
Preface
The mathematical sciences are part of nearly all aspects of everyday life.
The discipline has underpinned such beneficial modern capabilities as
internet searching, medical imaging, computer animation, weather predic-
tion, and all types of digital communications. Mathematics is an essential
component of computer science. Without it, you would find it challenging
to make sense of abstract language, algorithms, data structures, or differen-
tial equations, all of which are necessary to fully appreciate how computers
work. In a sense, computer science is just another field of mathematics.
It does incorporate various other fields of mathematics, but then focuses
those other fields on their use in computer science. Mathematics matters
for computer science because it teaches readers how to use abstract lan-
guage, work with algorithms, self-analyze their computational thinking,
and accurately model real-world solutions. Algebra is used in computer
programming to develop algorithms and software for working with math
functions. It is also involved in design programs for numerical programs.
Statistics is a field of math that deploys quantified models, representations,
and synopses to conclude from data sets.
This book focuses on mathematics, computer science, and where the
two intersect, including heir concepts and applications. It also represents
how to apply mathematical models in various areas with case studies. The
contents include 29 peer-reviewed papers, selected by the editorial team.
xvii
1
A Comprehensive Review on Text
Classification and Text Mining Techniques
Using Spam Dataset Detection
Tamannas Siddiqui and Abdullah Yahya Abdullah Amer*
Department of Computer Science, Aligarh Muslim University, Aligarh, UP, India
Abstract
Text data mining techniques are an essential tool for dealing with raw text data
(future fortune). The Text data mining process of securing exceptional knowledge
and information from the unstructured text is a fundamental principle of Text data
mining to facilitate relevant insights by analyzing a huge volume of raw data in
association with Artificial Intelligence natural language processing NLP Machine
Learning algorithms. The salient features of text data mining are attracted by the
contemporary business applications to have their extraordinary benefits in global
area operations. In this, a brief review of text mining techniques, such as cluster-
ing, information extraction, text preprocessing, information retrieval, text clas-
sification, and text mining applications, that demonstrate the significance of text
mining, the predominant text mining techniques, and the predominant contem-
porary applications that are using text mining. This review includes various exist-
ing algorithms, text feature extractions, compression methods, and evaluation
techniques. Finally, we used a spam dataset for classification detection data and a
three classifier algorithm with TF-IDF feature extraction and through that model
achieved higher accuracy with Naïve Bayes. Illustrations of text classification as
an application in areas such as medicine, law, education, etc., are also presented.
Keywords: Text mining, text classification, spam detection, text preprocessing,

text analysis
*Corresponding author: ayaamer@myamu.ac.in
Sharmistha Ghosh, M. Niranjanamurthy, Krishanu Deyasi, Biswadip Basu Mallik, and Santanu Das (eds.)
Mathematics and Computer Science Volume 2, (1–18) © 2023 Scrivener Publishing LLC
1
2 Mathematics and Computer Science Volume 2
1.1 Introduction
Text data mining techniques are predominantly used for extracting relevant
and associated patterns based on specific words or sets of phrases. Text data
mining is associated with text clustering, text classification, and the prod-
uct of granular taxonomy, sentiment analysis, entity relation modeling, and
document summarization [1]. Prominent techniques in text mining tech-
niques include extraction, summarization, categorization, retrieval, and
clustering. These techniques are used to infer distinguished, quality knowl-
edge from text from previously unknown information and different written
resources obtained from books, emails, reviews, emails, and articles with
the help of information retrieval, linguistic analysis, pattern recognition,
information extraction, or information extraction tagging and annotation
[2]. Text preprocessing is the predominant functionality in text data min-
ing. Text preprocessing is essential to bring the text into a form that can be
predictable and analyzable for text data mining. Text preprocessing can be
done in different phases to formulate the text into predictable and analyz-
able forms. These are namely lowercasing, lemmatization, stemming, stop
word removal, and tokenization. These important text preprocess steps are
predominantly performed by machine learning algorithms for natural lan-
guage processing tasks. These preprocessing steps implement data cleaning
and transformation to eliminate outliers and make it standardized to cre-
ate a suitable model to incorporate the text data mining process [3]. Text
data mining techniques are predominantly used for records management,
distinct document searches, e-discovery, organizing a large set of a text data,
analysis and monitoring of understandable online text in internet commu-
nication and blogs, identification of large textual datasets associated with
patients during a clinical area, and clarification of knowledge for the readers
with more extraordinary search experience [4]. Text data mining techniques
are predominantly used in scientific literature mining, business, biomedical,
and security applications, computational sociology, and digital humanities
as shown in Figure 1.1 below.
Dimensionality
reduction
Features classification Predict test

extraction learn model data
Evaluate
model
Figure 1.1 Overview of text classification.

Review on Text Classification and TM Techniques 3
Table 1.1 Text classification compared model classifiers.

Model Features
classifiers Authors Architecture extraction Corpus
SVM and C. W. Lee Gravity Similarity Wikipedia
KNN et al. Inverse TF-IDF
[7] Moment vectorizer
Logistic L. Kumar Bayesian TF-IDF RCV1-v2
Regression et al. Logistic
[13] Regression
Naïve A. Swapna Weight Weights Reuters-31678
Bayes NB et al. [9] Enhancing words
Method
SVM T. Singh String TF-IDF 20
et al. Subsequence vectorizer Newsgroups
[11] Kernel
The paper reviews text data mining techniques, various steps involved
in text preprocessing, and multiple applications that implement text data
mining methods discussed in Table 1.1.
1.2 Text Mining Techniques

Text Mining (TM) indicates informational content involved in several
sources like newspapers, books, social media posts, email, and URLs. Text
data summary and classification are typical applications of text mining,
particularly among different fields. It is appropriate to discuss some of the
techniques applied to achieve them through the step set shown in Figure
1.2 below.
1.2.1 Data Mining

Text mining is empowered in big data analytics to analyze unstructured
textual data to extract new knowledge and distinguish significant patterns
and correlations hidden in the huge amount of data sets. Big data analytics
are predominantly used for extracting the information and patterns that
are hidden implicitly in the data sets in the form of automatic or semi-
automatic unstructured formats or natural language texts. To perform this
test, mining operations, unsupervised learning algorithms, and supervised
learning algorithms or methods are predominantly used. These methods’
Text
collection
Preprocessing
Text data
Feature Selection
Classification TM Techniques Clustering
Information Summarization Information

Extraction Retrieve
Machine Learning Algorithms Knowledge
Figure 1.2 Text data mining techniques.
functionality is used for classification and prediction by using a set of pre-

dictors to reveal hidden structures in the information database [5]. In this
process, text mining is performed using pattern matching on regular doc-
uments and unstructured manuscripts [6].
1.2.2 Information Retrieval

Information Retrieval [IR] is a prominent method in text data mining
techniques. The fundamental principle of IR is identifying documents
stored in the database in unstructured formats, which meets the require-
ments of the information needed from the large collection of documents
stored in the datasets. IR is available in three models: Boolean Model,
Vector Model, and Probabilistic Model. In text data mining techniques, IR
plays a vital role with the indexing system and collection of documents [7].
This method is predominantly used for locating a specific item in natural
language documents. IR is used for learned knowledge extraction to con-
vert text within structured data for interesting mining relationships [8]. It
has been identified as a big issue to discover the appropriate designs and
analyze text records from huge amounts of data. Text data mining tech-
nique IR has resolved the issue and successfully selected attractive patterns
from the greatest knowledge data sets. IR techniques are predominantly

used for choosing the appropriate text documents from the huge volume of
databases with enhanced speed within a short period. The text data mining
technique IR extracts the exact required text documents from the greatest
databases and presents the accuracy and relevance of results [9].
1.2.3 Natural Language Processing (NLP)

NLP linguistics is subfield of computer science and AI. The fundamental prin-
ciple of NLP is to deal with the connection between computer machines and
humans with an assistant of NLP to read, interpret, learn, and make sense of
languages spoken by humans in a valuable way. It is powered by AI, which can
facilitate the machines to read, understand, interpret, manipulate, and derive
meaning from human languages [10]. It is a prominent AI technology used
in text data mining to transform the unstructured text depicted in documents
and databases into normalized, structured data suitable for performing anal-
ysis or implementing machine learning algorithms [11]. Long Short-Term
Memory [LSTM] is one of the predominant AI Machine Learning algorithms
to remember values with a recurrent neural network’s help. Seq2seq model is
another predominant model used in the NLP technique, which works with
encoder-decoder structure. In this model, it initially built the vocabulary list
to identify the correct grammar syntax. It works with some tags to identify
the structured and unstructured language identified in the documents. The
named entity recognition model is another predominant model to identify
relevant names and classify names by their entity. It is used to find the names
of people, names of places, and any other important entity in the given dataset
in text or documents. The NLP process features a Preferences’ Graph [12]. It is
utilized to build a set of user preferences. While the document is written, the
repetitively chosen tense, adjectives, conjunctions, and prepositions are iden-
tified and NLP creates a User Preference Graph. Based on the graph, it pre-
dicts the next word of the sentence to calculate the probability theory. Word
embedding is another inbuilt mechanism obtained from the feature training
and language model of NLP. The terms and idioms are planned into a vector
of the actual number of graphs during this process. Th e model is designed
from big unstructured text to generate the most probable output from the
input text searched from the documents’ database [13].
1.2.4 Information Extraction

Information Extraction [IE] is the beginning point for a computer to
estimate unstructured language documents. IE is predominantly used to
Unstructured Structured
Web Text Sequences
Sign of the Zodiac:
1. Aries
2. Taurus
The second sign of the Zodiac is 3. Gemini...
Taurus. Most Common Cause of
Strokes are the third most common cause of Death in America:
1. Heart Disease
death in America today. 2. Cancer
No study would be complete without 3. Stroke...
mentioning the largest rodent in the Largest rodent in the
world, the Capybara. world:
1. Capybara
2. Beaver
3. Patagonian Cavies
Figure 1.3 Information extraction.
distinguish the important relationships involved in the document as shown

in Figure 1.3. It is working to distinguish the predefined arrangements in a
text with the pattern matching method. Information that cannot be utilized
for mining is processed and evaluated by Information Extraction when the
documents consist of information. IE is rich with a fundamental mecha-
nism to distinguish the unstructured text available in the articles, blogs,
emails, reviews, and other documents with predefined arrangements for
post-processing. Post-processing is essential for performing web mining
and searching tools. IE is associated with the quality of extracting knowl-
edge from text. IE performs the extracting knowledge from text operations
from the unstructured text instead of abstract knowledge. IE performs text
mining tasks and further methods to explore the information from the text
data in hand [14].
1.2.5 Text Summarization

Text summarization is the part of text data mining techniques used to dis-
till the predominant data from a source to generate an abridged version
for a specific user and related task. The process of summarization is a step-
by-step process. Step one is converting the paragraph into sentences. Step
two is text processing. Step three is tokenization. Step four evaluates the
weighted occurrence frequency of the words and finally, step five substi-
tutes the words with the respective weighted frequencies. The process of
summarization occurs in two ways. These are extractive summarization
and abstractive summarization. The fundamental utility of summariza-
tion is to reduce the text’s content while preserving the meaning of the
text. Abstractive summarization is a technique the summary produces by
generating new sentences by rephrasing or using novel words. The process
finally extracts the important sentences from the document. The extractive
summarization extracts the important information or sentence from the

given text document or original document. A statistical method is needed
to be implemented on the original document [19].
1.2.6 Text Categorization

The fundamental objective of text categorization in text mining is to sort
the documents into groups. This can be done automatically by employing
the technique to perform the classification using natural language pro-
cessing and machine learning. This process categorizes consumer reviews,
customer support tickets, blogs, complaints, and other content-rich text
documents [20]. Text categorization has several features made from attri-
butes of the documents. Predominantly, the features are developed from
words and broader content of the corpus for higher dimensional features.
Text categorization plays a vital role in performing news classification,
sentiment analysis, and web page classification. Text categorization can
be implemented along with Linear Support Vector Machine algorithms to
achieve the best results [21].
1.2.7 Clustering
In clustering, the main objective is to perform cluster analysis of content-
based text records. It uses ML and NLP to recognize and classify unstruc-
tured text data using extracted descriptors from a targeted document
available in the database. Word clustering is part and parceling of text
clustering for part sets of information in subsets of semantically related
words. It is aimed to perform duties ranging from word sense or basic dis-
ambiguation to knowledge retrieval [22]. Unsupervised text clustering is
widely used in NLP to group similar data. It obtains the distance between
the points. The predominant clustering methods are namely soft clustering
and hard clustering. Hard clustering performs the grouping task for every
object that belongs to exactly one cluster. Hard clustering allows an object
to be grouped with one or more clusters based on its meaning and nature
[23].
1.2.8 Information Visualization

Information Visualization is meant for information analysis and enables
knowledge discovery via interactive graphical representations of textual
data. It enables exploration and understanding. It can be formulated in
the form of a tag or a word with varied size, position based on frequency,
categorization, or importance of the tag. Information visualization or

data visualization can be performed by employing seeing factors as maps,
graphs, and charts to understand the pattern extracted from text data.
These main objectives of text visualization are used to summarize large
amounts of text, formulate text data that is easy to understand, identify
the insights in qualitative data, and discover hidden trends and patterns of
the text. Text visualization can be done with the help of machine learning.
This can be performed to make sense of qualitative data quickly, easily, and
at scale. Word clouds are a great initiation for visualizing qualitative data.
World clouds facilitate insights useful for exploratory analysis to learn the
insights of the dataset and define labeling criteria for more advanced text
analysis. AI Machine Learning algorithms are used for sorting the data into
categories. Text visualization enables the extraction of the data’s actionable
insights and reveals the trends of the data [24].
1.2.9 Question Answer

Question answering is a subfield of information retrieval and NLP linguistic
computer science is concerned with creating systems in order to automat-
ically answer questions modeled through human language. Implementing
machine learning algorithms for the process of Question/Answer in a full
data-driven architecture can produce a practical solution for answering
the question with the right short answer rather than giving a list of pos-
sible answers. This can be implemented in the NLP to generate an accu-
rate answer with the knowledge database’s text similarity. Implementing
machine learning algorithms can produce the administration of the doc-
uments, analyze profiles and phrases, and perform marketing activities to
generate alerts to post the company branding information. A QA system
can be built with the help of python code and facilitate the user to post
the question, developing the internal mechanism to extract the question
and search for the matching documents with approximate string-matching
function from the knowledge base and extract the exact answer for the
question as shown in Figure 1.4 below [25].
key words extraction
User Interface Answering handel
Figure 1.4 Question answering system.

1.3 Dataset and Preprocessing Steps

In this review, we are utilizing a ‘spam_ham_dataset’ from Kaggle. This
is data set contains a total of 5172 documents with four columns. In the
dataset, emails were considered spam (1) or not (0), i.e., unwanted business
e-mail. The dataset is split into two parts: 70% is the training set and 30%
is the testing set. The experiment was conducted on the required dataset of
both the training set and testing set [15].
1.3.1 Text Preprocess

Text preprocessing is predominant in text data mining for dimensionality
reduction. Once the text is available in the knowledge database, it should
be preprocessed to implement the Machine Learning model. It is essen-
tial to perform preprocessing on the data with essential steps, namely
Tokenization, Lower Casing, Filtering, Stemming, and Lemmatization
[16]. The documents available in the knowledge database are described
as a vector in a multi-dimensional area. Every single word has a unique
dimension discussed below in Table 1.2 [33].
1.4 Feature Extraction

Text feature extraction involves carrying out a dictionary of terms from
the textual data then converting them into a feature set available to the
Table 1.2 Preprocessing steps.

Preprocessing Description
1. Tokenization Is the method of dividing the sentence within terms [26]
2. Filtering Token filtering is the process of filtering out any tokens that
are not useful for application. The token filtering process
eliminates the digits, punctuation marks, stop-word
tokens, and other unnecessary tokens in the text [27].
3. Lemmatization Lemmatization is the process of resolving the term to
Lemma, meaning the part of speech of the term.
Lemmatization transforms the word into a proper
root form with a part of the speech tagger [28].
4. Stemming Is the method used for transforming a term to its root;
this process is associated with the normalization task
associated with bluntly removing word affixes [29]
classifier. Next, we will show some techniques that can be applied to extract
features from text data.
1.4.1 Term Frequency – Inverse Document Frequency

A difficulty with the BoW method is that the terms with high-frequency dis-
play control in the data cannot provide much information about the model.
Also, due to that difficulty, control-specified words that have a lower score
may be ignored. To resolve that difficulty, the frequency of words is rescaled
by analyzing where the words frequently happen compared to the total text
document. Here, we applied TF-IDF. Term Frequency measures frequency
of the word in a popular text and the Inverse Document Frequency IDF
number represents words within the whole text document [30].
1.4.2 Bag of Words (BoW)

BoW is common among different feature extraction techniques and it
forms a word as the port feature set of total word instance. It is then recog-
nized as a “bag” of words for a process that does not mind the orders of the
words or how many times a word occurs, but just represents words already
within a record of words discussed below in Table 1.3 [31].
Table 1.3 Advantages and disadvantages of feature extraction.
Model Advantages Disadvantages
TF-IDF Simple to calculate BoW does not catch
Simple to divide the similarity situations in the text
among recognized TFIDF (syntactic)
Easy to extract a descriptive word BoW does not catch purpose
within the text document meaning in the text
General items do not impact an (semantics)
outcome because of IDF (for
example, “are”, “is”, etc.)
BoW Simple to compute It does not capture the place
Simple to count a similarity in the text (syntactic)
among documents using the • BoW does not capture
BoW purpose in the text
Easy extraction and metric (semantics)
descriptive words by a • Common terms affect
document outcomes (for example,
Task within unknown terms “am”, “is”, etc.)
1.5 Supervised Machine Learning Classification

In this part, we summarize existing text dataset classification algorithms.
We define the KNN [17] algorithm which is utilized for data classification.
Then, we describe other methods like logistic regression and Naïve Bayes
[18], which are commonly utilized in the scientific community as classifi-
cation techniques. Classification algorithms associated with the linear sup-
port vector machine have achieved high accuracy in the classification of
data and performance evaluation text data mining operations.
1.6 Evaluation
The experimental evaluation of text classifiers measures efficiency (i.e.,
capacity to execute the correct classification). Accuracy, recall, and preci-
sion are generally employed to measure the effectiveness of a text classifier.
Accuracy (FP+FN/TP+TN+FP+FN = 1-accuracy), on the other hand, is
not w idely used for text classification applications because it is insensitive
to variations in the number of correct decisions due to the large value of
the denominator (TP + TN), as discussed below in Table 1.4.
1.7 Experimentation and Discussion Results for

Spam Detection Data
In the experimental part of the review study, we used spam data to classify
files as either spam or non-spam by using machine learning algorithms.
Our experimental was done on the Windows system in a Python virtual
environment with Anaconda Jupyter to implement the code to build a
machine learning model. These steps are executed in five phases. The first
phase is to import the library, the second is to load and preprocess the
Table 1.4 Confusion matrix in machine learning.

Accuracy TP+TN/TP+TN+FP+FN
Recall TP/TP+FN
Precision TP/TP+FP
F1-Measure 2(P*R)/(P+R)
dataset, the third phase uses TF-IDF feature extraction techniques, the
fourth phase is to train and test our model (after dividing the dataset into
two parts: 70% training set and 30% testing set to fit the model), and the
last phase is to predict and evaluate the model by using amusement preci-
sion, recall, accuracy, and f score confusion matrix, as shown in Table 1.5
and Figure 1.5 below.
Table 1.5 shows the outcome of three classifier models. Through accu-
racy, recall, precision, and F1-Measure, Naïve Bayes achieved the higher
accuracy at around 97.06% among classifier models for detection of spam
files from a data set.
As shown in the Figure 1.5, Naïve Bayes achieved higher results and
higher improvement accuracy with a data trained spam model using
K-Nearest Neighbor and Logistic Regression algorithms.
Table 1.5 Spam dataset classification.

Accuracy Recall Precision
Algorithm used (%) (%) (%) F1-measure
Naïve Bayes 97.06 97.76 95.65 97.78
K-Nearest Neighbor 93.12 92.55 92.01 93.09
Logistic Regression 94.89 91.32 93.21 94.76
Spam Dataset classification
98
97
96 Naïve Bayes
95 Knearest Neighbor
94
Logistic Regression
93
92
91
90
89
88
Accuracy (%) Recall (%) Precision (%) F1‐Measure
Figure 1.5 Spam dataset classification.

1.8 Text Mining Applications

Text data mining is predominantly used in different applications. Text
mining’s fundamental objective is information extraction, information
retrieval, categorization, clustering, and summarization. These activities
are helping business applications with extensive operational capacity. The
main applications have already implemented text data mining techniques.
These are namely risk management applications, knowledge management
applications, content enrichment applications, business intelligence appli-
cations, contextual advertising applications, customer care services appli-
cations, cybercrime prevention applications, spam filtering, data analysis
from social media, and fraud detection through claims investigation appli-
cations, as shown in Figure 1.6 below [32, 34].
1.9 Text Classification Support

1.9.1 Health
Most text data in the health and medical domain displays unstructured
patient information by vague terms with typographical errors. Here, the
role of text classification process data and dealing with it in order to pre-
dict illness uses machine learning algorithms. The medical field consists of
selecting medical diagnoses to special class values taken from many cate-
gories and this is a field of healthcare where text classification techniques
can be deeply valuable [24].
Risk
Management
Fraud Customer
Detection Applications Care Services
of Text
Mining
Business Social Media

Intelligence Analysis
Figure 1.6 Applications of text data mining [32].

1.9.2 Business and Marketing

Profitable organizations and companies use social networks such as
Facebook and Twitter for marketing purposes such as shopping, buying,
selling, etc. Through mining and sentiments analysis, businesses know
what a customer desires about the producer in order to improve and
increase product from the producer. Additionally, text classification in
organizations and companies is an important tool for businesses to obtain
more customers efficiently.
1.9.3 Law
Government institutions have generated large volumes of legal record text
documents. Analysis and retrieving that data manually is so difficult. Here,
a system is required to deal with information and process it automatically
in order to help lawyers and their clients [22]. Organization of those record
text documents is the foremost difficulty to the law community. Building a
system to classify documents is helpful for the law community.
1.10 Conclusions
In this review paper, we present a review of text data mining techniques with
relevant descriptions. Text data mining techniques are applied to extract
readable knowledge from raw text data sets regularly into the unstructured
data. So, in this paper we presented the gist of various previous research
works presented on the predominant text mining techniques, namely Data
Mining, Text Classification, Information Extraction, Question Answering,
Topic Tracking, Natural Language Processing, Information Retrieval,
Text Summarization, Text Categorization, Clustering, and Information
Visualization. The paper has presented the important mechanism of
preprocessing text data mining with four steps: Tokenization, Filtering,
Lemmatization, and Stemming. We have discussed the process of text data
mining by presenting the fields that use text mining applications. Lastly,
we compared the most popular text classification algorithms. Finally, we
used a spam dataset for classification detection data and a three classifier
algorithm with TF-IDF feature extraction, achieving higher accuracy with
Naïve Bayes around 97.06%. Illustrations of the usage of text classification
as support for applications in medicine, law, education, etc. are related in
a separate part.
References
1. S. S. Kermode and V. B. Bhagat, “A Review: Detection and Blocking Social
Media Malicious Posts,” Int. J. Mod. Trends Eng. Res., vol. 3, no. 11, pp. 130–
136, 2016, doi: 10.21884/IJMTER.2016.3133.Q4M8O.
2. M. Allahyari et al., “A Brief Survey of Text Mining: Classification, Clustering
and Extraction Techniques,” 2017, [Online]. Available: http://arxiv.org/
abs/1707.02919.
3. C. C. Aggarwal and C. X. Zhai, “Mining text data,” Min. Text Data, vol.
9781461432, pp. 1–522, 2013, doi: 10.1007/978-1-4614-3223-4.
4. M. A. Zende, M. B. Tuplondhe, S. B. Walunj, and S. V. Parulekar, “Text min-
ing using python,” no. 3, pp. 2393–2395, 2016.
5. N. F. F. Da Silva, E. R. Hruschka, and E. R. Hruschka, “Tweet sentiment
analysis with classifier ensembles,” Decis. Support Syst., vol. 66, pp. 170–179,
2014, doi: 10.1016/j.dss.2014.07.003.
6. A. Nielsen, “Python programming — text and web mining,” Aust. J. Plant
Physiol., vol. 21, pp. 507–516, 2011, [Online]. Available: http://www2.imm.
dtu.dk/pubdb/views/edoc_download.php/5781/pdf/imm5781.pdf.
7. C. W. Lee, S. A. Licorish, B. T. R. Savarimuthu, and S. G. Macdonell,
“Augmenting text mining approaches with social network analysis to under-
stand the complex relationships among users’ requests: A case study of the
android operating system,” Proc. Annu. Hawaii Int. Conf. Syst. Sci., vol.
2016-March, pp. 1144–1153, 2016, doi: 10.1109/HICSS.2016.145.
8. N. U. Pannala, C. P. Nawarathna, J. T. K. Jayakody, L. Rupasinghe, and K.
Krishnadeva, “Supervised Learning-Based Approach to Aspect Based
Sentiment Analysis,” 2016 IEEE Int. Conf. Comput. Inf. Technol., pp. 662–
666, 2016, doi: 10.1109/CIT.2016.107.
9. A. Swapna, K. G. Guptha, and K. Geetha, “Efficient Approach for Web Search
Personalization in User Behavior Supported Web Server Log Files Using Web
Usage Mining,” vol. 2, no. 4, pp. 189–196,
10. G. Barbier, “Social Network Data Analytics,” 2011, doi: 10.1007/978-1-
4419-8462-3.
11. T. Singh, M. Kumari, T. L. Pal, and A. Chauhan, “Current Trends in Text
Mining for Social Media,” Int. J. Grid Distrib. Comput., vol. 10, no. 6,
pp. 11–28, 2017, doi: 10.14257/ijgdc.2017.10.6.02.
12. A. Shahrestani, M. Feily, R. Ahmad, and S. Ramadass, “Architecture for
applying data mining and visualization on network flow for botnet traffic
detection,” ICCTD 2009 - 2009 Int. Conf. Comput. Technol. Dev., vol. 1,
pp. 33–37, 2009, doi: 10.1109/ICCTD.2009.82.
13. L. Kumar and P. K. Bhatia, “Available Online at www.jgrcs.info TEXT
MINING: CONCEPTS, PROCESS, AND APPLICATIONS,” vol. 4, no. 3,
pp. 36–39, 2013.
14. K. N. S. S. V Prasad, S. K. Saritha, and D. Saxena, “A Survey Paper on Concept

Mining in Text Documents,” Int. J. Comput. Appl., vol. 166, no. 11, pp. 7–10,
2017.
15. p. Anitha and C. Guru Rao, “ Email Spam Filtering Using Machine Learning
Based Xgboost Classifier Method,” Int. J. Adv. Comput. Sci. Appl., Vol.12
No.11 (2021), pp. 2182-2190, doi: 10.14569/IJACSA.2015.060121.
16. N. T. and A. M., “Investigating Crimes using Text Mining and Network
Analysis,” Int. J. Comput. Appl., vol. 126, no. 8, pp. 19–25, 2015, doi: 10.5120/
ijca2015906134.
17. M. Zareapoor and S. K. R, “Feature Extraction or Feature Selection for Text
Classification: A Case Study on Phishing Email Detection,” Int. J. Inf. Eng.
Electron. Bus., vol. 7, no. 2, pp. 60–65, 2015, doi: 10.5815/ijieeb.2015.02.08.
18. D. Saxena, S. K. Saritha, and K. N. S. S. V. Prasad, “Survey on Feature Extraction
methods in Object,” Int. J. Comput. Appl., vol. 166, no. 11, pp. 11–17, 2017.
19. S. A. Salloum, M. Al-Emran, A. A. Monem, and K. Shaalan, “A Survey of
Text Mining in Social Media: Facebook and Twitter Perspectives,” Adv. Sci.
Technol. Eng. Syst. J., vol. 2, no. 1, pp. 127–133, 2017, doi: 10.25046/aj020115.
20. M. Neeraja and J. Prakash, “Detecting Malicious Posts in Social Networks
Using Text Analysis,” vol. 5, no. 6, pp. 2015–2017, 2016.
21. G. S. L. Vishal Gupta, “A Survey of Text Mining Techniques and Applications,” J.
Emerg. Technol. web Intell., vol. 1, no. 1, p. 17, 2009, doi: 10.4304/jetwi.1.1.60-76.
22. M. Sukanya and S. Biruntha, “Techniques on text mining,” Proc. 2012 IEEE
Int. Conf. Adv. Commun. Control Comput. Technol. ICACCCT 2012, no.
978, pp. 269–271, 2012, doi: 10.1109/ICACCCT.2012.6320784.
23. T. Siddiqui, A. Y. A. Amer, and N. A. Khan, “Criminal Activity Detection
in Social Network by Text Mining: Comprehensive Analysis,” 2019 4th Int.
Conf. Inf. Syst. Comput. Networks, ISCON 2019, pp. 224–229, 2019, doi:
10.1109/ISCON47742.2019.9036157.
24. A. Yahya, A. Amer, and T. Siddiqui, “Detection of Covid-19 Fake News text
data using Random Forest and Decision tree Classifiers Abstract:,” vol. 18,
no. 12, pp. 88–100, 2020.
25. M. R. Begam, “Survey: Tools and Techniques implemented in Crime Data
Sets,” vol. 2, no. 6, pp. 707–710, 2015.
26. M. Ramageri, “DATA MINING TECHNIQUES AND APPLICATIONS,”
vol. 1, no. 4, pp. 301.
27. G. Nandi and a. Das, “A Survey on Using Data Mining Techniques for Online
Social Network Analysis.,” Int. J. Comput. Sci. Issues, vol. 10, no. 6, pp. 162–167,
2013, [Online]. Available: http://search.ebscohost.com/login.aspx?direct=
true&profile=ehost&scope=site&authtype=crawler&jrnl=16940784&AN=
93404019&h=MWYzsNVeP2Q8klAFiWFHW3PUpgJLZxRIpB1jfSK4qJfB-
baMUEp4nY/oJdYPRHc4xHL0dBYfuGxhZsmiP7ToLBg==&crl=c.
28. T. Siddiqui, N. A. Khan, and M. A. Khan, “PMKBEA: A Process Model Using
Knowledge Base Software Engineering Approach,” pp. 5–7, 2011.
29. S. Alami and O. Elbeqqali, “Cybercrime profiling: Text mining techniques to

detect and predict criminal activities in microblog posts,” 2015 10th Int. Conf.
Intell. Syst. Theor. Appl. SITA 2015, 2015, doi: 10.1109/SITA.2015.7358435.
30. “Build Your First Text Classifier in Python with Logistic Regression | Kavita
Ganesan.” https://kavita-ganesan.com/news-classifier-with-logistic-regression-
in-python/#.X4xyddD7Q7e (accessed Oct. 18, 2020).
31. V. Kumar and L. Velide, “A DATA MINING APPROACH FOR PREDICTION
AND TREATMENT Supervised machine learning algorithm:” vol. 3, no. 1.
pp. 73–79, 2014.
32. P. C. Thirumal and N. Nagarajan, “Utilization of data mining techniques
for the diagnosis of diabetes mellitus - A case study,” ARPN Journal of
Engineering and Applied Sciences, vol. 10, no. 1. pp. 8–13, 2015.
33. E. M. G. Younis, “Sentiment Analysis and Text Mining for Social Media
Microblogs using Open Source Tools: An Empirical Study,” Int. J. Comput.
Appl., vol. 112, no. 5, pp. 44–48, 2015, doi: 10.1093/ojls/gqi017.
34. Amer A.Y.A., Siddiqui T. (2021) Detecting Text-Bullying on Twitter Using
Machine Learning Algorithms. In: Bhattacharya M., Kharb L., Chahal D.
(eds) Information, Communication and Computing Technology. ICICCT
2021. Communications in Computer and Information Science, vol 1417.
Springer, Cham. https://doi.org/10.1007/978-3-030-88378-2_17
2
Study of Lidar Signals of the
Atmospheric Boundary Layer
Using Statistical Technique
Kamana Mishra1* and Bhavani Kumar Yellapragada2
School of Mathematical and Statistical Sciences, Indian Institute of Technology

1
Mandi, Himachal Pradesh, India

2
Department of Space, National Atmospheric Research Laboratory Tirupati, India
Abstract
The rotational turbulence caused by mixing the layers of air, wind shear compo-
nents, mountain waves, aerosol particles, and other pollutants affects the lowest
and densest layer of the earth’s surface troposphere. Due to the turbulence, the
height of the convective boundary layer (CBLH) changes over the day dramat-
ically. We observe the variation in peak positions of lidar backscatter signals by
performing a statistical technique for analyzing the behavior of the convective
boundary layer (CBL). After that, to examine the behavior of the whole boundary
layer, a distribution method and histogram plots will be used. We provide the sta-
tistical method for getting the best fit distribution to show how the result leads to
the physical observation of data.
Keywords: Statistical technique, lidar data, distribution plots, convective

boundary layer
2.1 Introduction
Surface forcing of the planetary boundary layer influences the closest and
deepest part of the earth’s atmosphere, i.e., the troposphere. The turbu-
lence in the troposphere causes the high variability of the temperature,
moisture, wind shear, and pollutant particles. Due to surface friction,
*Corresponding author: kamana182mis@gmail.com
19
winds in ABL are weaker than above and tend to blow towards the area
where pressure is low. It is noticeable over the Indian subcontinent that the
height of the convective boundary layer (CBLH) is low during the winter
and monsoon and high in the sunny days as the warmer lower-layer air
mixes with the cool air and convection arises in the summertime only.
This behavior indicates that solar heating affects the CBLH, and hence,
CBL is also named the daytime planetary boundary layer. The convec-
tion occurs by mixing of airs forming bubbles of warmer air or eddies.
The small eddies generated by the local wind shear cause turbulence in
the surface layer and have a completely random behavior, however, the
turbulence in the mixed layer is caused by large bubbles of warmer air,
and hence, it is not completely random. Since the turbulent processes are
found to be nondeterministic, it is beneficial to use a robust technique
like statistical methods to find out the variability in the boundary layer
and its key parameter as CBLH and has been studied by Pal, S., Behrendt,
A., and Wulfmeyer V. in Elastic-backscatter-lidar-based characterization
of the convective boundary layer and investigation of related statistics [3].
The signal which will be used in the study of CBLH is the total backscat-
ter from the atmosphere that represents combined aerosol and molecular
backscatter as mentioned earlier by Dang R, Yang Y, Hu X-M, Wang Z,
and Zhang S [2]. For using the statistical method, we need the informa-
tion in terms of data and we choose an instrument with high resolution
power named Lidar (Light Detection and Ranging), which is a remote
sensing method used in measuring the ranges to the earth and the concept
of using Lidar to detect ABL height relies on the assumption that there is
a strong gradient in the concentration of aerosols in the CBL versus the
free atmosphere. The importance of defining a new technique is that this
method works for single scan data and uses spatial average, however, the
variance method needs multiple numbers of scans to identify the variance
and use time average.
After finding the CBLH through statistical technique, we will try to
examine the behavior of the whole boundary layer rather than just the CBL
part. The histogram plots and distribution method can serve this purpose
very well, meaning a list of best fit distribution along with each signal can
be used to examine the behavior of the boundary layer, used before by the
authors in ‘The Determination of Aerosol Distribution by a No-Blind-Zone
Scanning Lidar” [6]. After getting the list of all possible distributions, it will
be checked whether there is any range of the signals which follow a speci-
fied distribution or if the signals are following the distributions randomly.
Now, to find the distribution of the backscatter signals manually for any
structure of the boundary layer, a new technique needs to be introduced
Study of Lidar Signals of the ABL Using Statistical Technique 21
which holds in practical life. In this new technique, we will try to use the
statistical approach. Lastly, we check the consistency for applying both the
statistical techniques by using two different datasets and relate the math-
ematical result with the physical observations to examine the behavior of
ABL.
2.2 Methodology
2.2.1 A Statistical Approach to Determine the CBLH
Due to the high-resolution power of Lidar, signals of Lidar can be used in
finding the CBLH as Georgoulias, A.K., Papanastasiou, D.K., and Melas, D.
et al. used in “Statistical analysis of boundary layer heights in a suburban
environment” [4]. There are different types of Lidar techniques that serve a
specific purpose, such as temperature profile using vibrational Raman (VR)
and rotational Raman (RR). Backscattering Lidar and Elastic-backscatter
Lidar (EBL) were used in ‘New Technique to Retrieve Tropospheric
Temperature Using Vibrational and Rotational Raman Backscattering’ [5].
In our statistical technique, we do simple differentiation of the signal con-
cerning the height, i.e., tracking the local maxima points, then we do some
restrictions on the points of local maxima to indicate the real position of
peaks of backscatter signal. The peak position changes with change in val-
ues of the restriction set. The peak position changes with change in values
of the restriction set given below in Table 2.1.
Table 2.1 Restriction set used in algorithm for detecting peak position of lidar
signals.
Elements of set of
restriction Meaning Default value
mph Minimum peak height None
mpd Minimum distance between 1 (positive integer)
two peaks
threshold Noise level 0
kpsh Keep peaks with the same False
height even if they are closer
than `mpd`
valley Local minima False
2.2.2 A Statistical Approach to Determine the Best

Fit Distribution to the Backscatter Signals of the
Lidar Dataset
In this technique, we start with the histogram plots which help in decid-
ing which type of distributions can take place and after that, a list of
distributions will be prepared, which can be discrete or continuous.
Now, to get the best fit distribution one needs to use tests under the null
hypothesis to prove that this particular distribution holds well. There are
many tests like the chi-square test for independence, ANOVA (analysis
of variance), Homogeneity of Variance (HOV), Mood’s Median, Distance
Covariance Test (d cov), Kolmogorov-Smirnov test (KS test), Welch’s
T-test, Kruskal-Wallis H Test, etc. that are available to use but choos-
ing a test is a crucial part of the procedure. So, after roughly looking
at the histogram plots of the data and examining the previous research
defined in “Analysis of ocean clutter for wide-band radar based on real
data” [7], “A Speckle Filtering Method Based on Hypothesis Testing for
Time-Series SAR Images” [8], “Structural Dynamics of Tropical Moist
Forest Gaps” [9], and “Non-cooperative signal detection in alpha stable
noise via Kolmogorov-Smirnov test” [10], it can be observed that the
data follows continuous distribution and since the Kolmogorov-Smirnov
test (KS test) can be applied only for the data which follows continuous
distribution and has a limitation that the distribution must be fully spec-
ified, the k-s test is the best option according to our dataset. Now, after
defining the list of possible distributions and choosing the test, the per-
centile bins will be defined so that we can make use of a particular range
of the data rather than the exact dataset. An increment in the number of
percentile bins results in a more accurate distribution type. That is why
we check whether the result holds for a large number of bins or not. After
that, observed frequency and expected frequency is calculated to formu-
late the chi-square statistics. The fitness of distributions will be sorted
based on p-value, which is calculated from the hypothesis and the chi-
square statistics. Since all the parameters are defined here, to use this test
for a large number of columns a function can be created and by giving
the number of columns, one can get the name of the best fit distribution.
Now, we can plot the histogram with the best-fit distribution curve to
check whether our formulation of function holds or not. One can also
get distribution with its value of scale and shape parameters using the
format function which will be helpful to make a graph of the distribution
corresponding to its parametric values.
2.3 Mathematical Background of Method

This new technique of finding CBLH is defined on the behalf of mathemat-
ics and it will pave the way for using the statistical methods more often in
the future for observing the behavior of atmospheric data. To use the statis-
tical approach, a dataset of backscatter signals is required and Lidar helps
us get this dataset. Detecting the peaks of the signals implies getting a set of
those points from the Lidar dataset, at which there is a steep gradient and
differentiation serves this purpose very well. Since the values of this dataset
are completely random and defining a function is not possible at all, we find
the points of local maxima through the graph. The slope is zero at the points
where there is local maxima or minima. Let us define this collection of local
maxima and local minima points by A and set of local maxima points by
B. Now, let a ∈ A. If the slope is positive ∀ y ∈ (a-δ, a) and negative ∀ y ∈
(a, a+ δ), then a ∈ B and if the slope is negative ∀ y ∈ (a-δ, a) and positive ∀
y ∈ (a, a+ δ), then a ∈ A\B. After getting elements of B, we will move to the
next step which means checking whether the point of local maxima is peak
or not because the first derivative has a downward zero-crossing at peak
maximum. But, in real-life experiments there is some random noise that
produces the false peak points and the result will be affected if we include
these false positions of peaks also. To avoid this problem, one has to put
some restrictions which are defined earlier in the methodology.
After that, a statistical technique will be used to find the distribution
which fits best to the given data. The histogram is a helpful tool to figure
out what kind of distribution data we have. Hence, by roughly analyzing
the histogram plot, one can define a set of possible distributions (continu-
ous or discrete) which can fit the data and choose the test accordingly. We
will make use of the KS test here, which decides whether the sample comes
from a population with specific distribution and is based on the empirical
distribution function (ECDF). If N data points are given, then firstly order
them from smallest to largest value, which can be denoted as X1, X2, …,
XN. Then, ECDF can be defined as:
EN = n(i)/N (2.1)
where n(i) is the number of points that are less than Xi.
K-S Test:
(Null hypothesis) H0: data follows given distribution
(Alternative hypothesis) H1: data does not follow the given dis-
tribution function
Test Statistic:
This test is defined as:
 i −1 i 
=D max  F ( Xi ) − , − F ( Xi )  (2.2)
1≤i ≤N  N N 
where F is the cumulative distribution function of the distribution that will

be tested with the condition that it must be continuous distribution, as well
as fully specified, which means parameters (such as local, shape, and scale)
cannot be calculated from the dataset.
Significance level: α
Critical Value: We reject the null hypothesis if the value of D is greater
than the critical value obtained from the table and this value can be written
as a p-value.
So, a p-value based on the approximation of the distribution of the test
statistics under the null hypothesis is calculated. Based on the p-values and
chi-square statistics, we will sort the distribution starting from the best fit
distribution. Since all the parameters have been termed, we can define a
function using all these, send the value of the data column-wise, and then
get the desired best fit distribution. One remark needs to be stated here:
more than one distribution can be fit to the given pattern of data with a
very small difference in their p-values. In that case, we can take the top
three distributions and get a distribution that fits for some particular range
of columns. This analysis emphasizes the databases with various forms of
utterances in Tamil language for the applied classification technique and
the obtained accuracy with the technique used to extract the feature. The
number of samples which have been taken for this are tabulated in esca-
lating order.
2.4 Example and Result

We will now consider an example of the Lidar dataset of backscatter sig-
nals of the boundary layer and scans are taken for every second for the
time interval of ten minutes. In Figure 2.1, the temporal variation of Lidar
signals shows the rise of boundary layer thermals. We have taken here the
scans for a dataset in the interval 09:24-09:34. In ‘Analyzing the atmo-
spheric boundary layer using high-order moments obtained from multi-
wavelength lidar data: impact of wavelength choice’ the authors have made
×106
3 9
8
2.5 7
Range (km)
6
2
5
1.5 4
3
1 2
1
0.5
0
09:24 09:25 09:26 09:27 09:28 09:29 09:30 09:31 09:32 09:33
Time (IST)
Figure 2.1 Temporal variation of lidar signal for dataset-1.
the same figures for the larger time period from 13:00 to 19:00 UTC and
from 17:00 to 21:00 UTC [1]. The reason for taking this time interval is
that the turbulence in the CBL increases with the rise in sunrays, i.e., in
the morning and due to the temperature variation boundary layer thermal
increases. However, if we take the time-interval in the night, then turbu-
lence ceases with the decrease in the amount of solar heating.
After using the statistical technique for all the backscatter signals, it is
observed that after a certain height of signal, data behaves similarly for all
the backscatter signals, which means the point at which there is a steep
gradient approximately has the same location for each signal. This means
that after a certain height of the layer, less presence of aerosols and other
pollutant particle ceases the turbulence. Hence, a part of the backscatter
signal in which the turbulence is very high is used rather than using the
whole range. After getting the peak points for each scan, we make a graph
of peak points along with scan number for a time-interval.
Now, if we compare these two figures, then it can be seen that the graph
of peak points detected using the statistical technique matches the figure
containing the rise of boundary layer thermals. When one observes graphs
of the peak points for each signal in the included excel sheet, it can be eas-
ily seen that in starting there is a peak point near the first peak and after
some signals, there is no point of the peak near the first peak point. Again,
there is a peak point found near the first one which shows the dissimilarity
in the location of peaks of the backscatter signals. From Figure 2.2, one can
notice that peak points are randomly distributed over the interval. Also,
the changes in peak height are not following any particular pattern. This
result shows that the changes in the part of the surface layer of CBL do
not follow any identified pattern and hence cannot be organized into any
predefined structure.
2000
Altitude (m)
1500
0924 0925 0926 0927 0928 0929 0930 0931 0932 0933
Local Time (LT)
Figure 2.2 Peak points with scan numbers for dataset-1.
However, after excluding a certain range, peak points attain the approx-
imately same position, which physically shows that the turbulence above
this shallow layer in the mixed layer is not completely random and can be
organized into some predefined structure like plums and thermals. This
method seems to be consistent and to confirm this we will take another
dataset with the time interval 09:34-09:44, having the same interval length.
The same process will be used here and after getting the peak points for
each scan, we will make a graph of peak points for dataset-2 along with
scan number. This method seems to be consistent and to confirm this we
will take another dataset with the time interval 09:34-09:44, having the
same interval length as shown in Figure 2.3. The same process will be used
here and after getting the peak points for each scan, we will make a graph
of peak points for dataset-2 along with the scan number as shown in Figure
2.4.
×106
3 9
8
2.5 7
6
Range (km)
2
5
1.5 4
3
1 2
1
0.5
0
09:34 09:35 09:36 09:37 09:38 09:39 09:40 09:41 09:42 09:43 09:44
Time (IST)
Figure 2.3 Temporal variation of lidar signal for dataset-2.

2000
Altitude (m)
1500
0934 0935 0936 0937 0938 0939 0940 0941 0942 0943
LOCAL TIME (LT)
Figure 2.4 Peak points with scan numbers for dataset-2.
In this, if we compare both the figures then the peak points clearly show
the temporal variation of Lidar signals. Hence, the result shows the consis-
tency of the statistical method used here and this method can be utilized in
detecting peak positions of backscatter signals.
After using the statistical technique (k-s test statistic), the distributions
which fit the dataset of the backscatter signals have been found and a list of
them along with the histogram plot, as well as the graph of best-fit distribu-
tion in an excel file are included in the link. Now, if we analyze the graphs
up to the two-thirds part of the dataset, gamma or beta distribution fits in
its maximum part, which has an almost same graph, and in the remaining
one-third of the dataset, t-distribution fits perfectly. However, a few data-
sets perform the log norm, chi-square, inverse Gaussian, and Pearson 3
distributions. If we use the same technique for dataset-2 and analyze the
graphs, the whole dataset follows t-distribution. This is probably happen-
ing because we have taken this dataset for the time-interval which is after
the time for the first dataset.
2.5 Conclusion and Future Scope

The observations we have obtained using a statistical technique for the scans
are the same as the temporal variation of Lidar signals. We have checked it
using another dataset. Hence, the method gives a consistent result and can
be utilized in detecting peak positions of backscatter signals. After that, we
move to analyze the behavior of the whole boundary layer rather than only
the CBL part using the distribution method for scans. Then, it is noticeable
that the changes in the distribution with some particular ranges of signals
are happening due to the physical process involved in that. Sometimes we
get a broader spectrum and sometimes a narrower spectrum. The reason
for this change is that the broader spectrum indicates turbulence process
and wind variability, whereas the narrower spectrum indicates tempera-
ture inversion and is associated with stable layer phenomenon. The fitting
of distribution is affected by thermal heating, water vapor condensation,
and other physical processes involved during the day. So, we have exam-
ined the natural environment and now this study leads us to examine the
changes in the manmade environment. The consistency of this method is
paving the way to use it in different fields such as hydrodynamic modeling,
storm surge modeling, etc.
Acknowledgement
I would like to express my gratitude to the team of Lidar project of NARL,
Tirupati, India for providing the data for the research work.
References
1. de Arruda Moreira, G., da Silva Lopes, F. J., Guerrero-Rascado, J. L., da
Silva, J. J., Arleques Gomes, A., Landulfo, E., and Alados-Arboledas, L.,
07 Aug 2019. Analyzing the atmospheric boundary layer using high-or-
der moments obtained from multi- wavelength lidar data: impact of wave-
length choice, Atmos. Meas. Tech., 12, 4261– 4276, https://doi.org/10.5194/
amt-12-4261-2019.
2. Dang R, Yang Y, Hu X-M, Wang Z, Zhang S, 2019. A Review of Techniques
for Diagnosing the Atmospheric Boundary Layer Height (ABLH) Using
Aerosol Lidar Data. Remote Sensing. 11(13):1590. https://doi.org/10.3390/
rs11131590.
3. Pal, S., Behrendt, A., and Wulfmeyer, V., 2010. Elastic-backscatter-lidar-
based characterization of the convective boundary layer and investigation
of related statistics, Ann. Geophys., 28, 825–847, https://doi.org/10.5194/
angeo-28-825-2010.
4. Georgoulias, A.K., Papanastasiou, D.K., Melas, D. et al., 2009. Statistical anal-
ysis of boundary layer heights in a suburban environment. Meteorol Atmos
Phys 104, 103– 111. https://doi.org/10.1007/s00703-009-0021-z.
5. Su, J., McCormick, M. P., & Lei, L. (2020). New Technique to Retrieve Tropo
spheric Temperature Using Vibrational and Rotational Raman Backscattering.
Earth and Space Science, 7. https://doi. org/10.1029/2019EA000817.
6. Wang, J.; Liu, W.; Liu, C.; Zhang, T.; Liu, J.; Chen, Z.; Xiang, Y.; Meng, X, 2020.
The Determination of Aerosol Distribution by a No-Blind-Zone Scanning
Lidar. Remote Sens. 12, 626. https://doi.org/10.3390/rs12040626
7. QingWeiPing, 2011. Analysis of ocean clutter for wide-band radar based on
real data https://doi.org/10.1145/2071639.2071669
8. Yuan, J.; Lv, X.; Li, R, 2018. A Speckle Filtering Method Based on Hypothesis
Testing for Time-Series SAR Images. Remote Sensing, 10, 1383. https://doi.
org/10.3390/rs10091383.
9. Hunter MO, Keller M, Morton D, Cook B, Lefsky M, Ducey M, et al. (2015)
Structural Dynamics of Tropical Moist Forest Gaps. PLoS ONE 10(7):
e0132144. https://doi.org/10.1371/journal.pone.0132144
10. J. Luo, S. Wang, E. Zhang and J. Luo, 2015. “Non-cooperative signal detection
in alpha stable noise via Kolmogorov-Smirnov test” 2015 8th International
Congress on Image and Signal Processing (CISP), 2015, pp. 1464-1468, doi:
10.1109/CISP.2015.7408114.
Annexure
Flowchart and Algorithm for peak detection
Flowchart:
START
Input mph, mpd
No
for j=1,
j≤=n,
Yes
Read data of jth column
x=[data], size = size(x), dx= x[1:]-x[:1]
No
for i=1,
j<size,
No
Yes
No Track if value
maxim >mph
Yes
ZK = value
Yes
ZK+1 - ZK
≥mpd
No
Yes
Exit Peak_value = value

Algorithm:
Step-1. Start
Step-2. Input the minimum peak height (mph) and minimum distance
between two peaks (mpd).
Step-3. Start a for loop to read columns of multidimensional data.
Step-4. Read the data of the column if ‘for’ loop is true else, go to step 11.
Step-5. Introduce a variable x to store the data in an array form. Then, use
the size function to find the size of x and after that, use the differentiation
function to differentiate the variable x.
Step-6. Again, start a ‘for’ loop for data stored in x.
Step-7. Track local maxima of data if for loop is true else, go to step 4 and
restart the process for the next column.
Step-8. If it is a point of maxima then, check whether its value is greater
than mph or not. If yes, go to step otherwise go to step 7.
Step-9. Introduce a new variable Zk and store the value in it.
Step-10. If difference between two local maxima i.e. Zk+1 - Zk is greater than
mpd then, assign this value as peak value else go to step 7.
Step-11. Exit.
Algorithm for finding the distribution:

Step-1. Start.
Step-2. Make a function in which column number will be passed.
Step-3. Define a variable which takes the data of the column and size of the
data.
Step-4. Input the list of possible distributions to find the best one among
them.
Step-5. Input the number of percentile bins.
Step-6. Calculate the observed and expected frequency.
Step-7. Calculate the p-value by using the predefined Kolmogorov-Smirnov
(KS) Test
Step-8. Calculate the value of chi-square statistics by using the formula as:
(Oi − ei )2
χ2 = ∑ Oi
where Oi and are observed and expected frequencies
respectively.
Step-9. Pass the number of columns in the function.
Step-10. Plot the histogram with the graph of best fit distribution for
cross-checking the result.
Step-11. Stop.
3
Optimal Personalized Therapies in Colon
Cancer Induced Immune Response
using a Fokker-Planck Framework
Souvik Roy* and Suvra Pal
Department of Mathematics, The University of Texas at Arlington, Arlington, TX, USA
Abstract
In this paper, a new stochastic framework to determine optimal combination thera-
pies in colon cancer-induced immune response is presented. The dynamics of colon
cancer are described through an Itö stochastic process, whose probability density
function evolution is governed by the Fokker-Planck equation. An open-loop control
optimization problem is proposed to determine the optimal combination thera-
pies. Numerical results with combination therapies comprising of the chemotherapy
drug Doxorubicin and immunotherapy drug IL-2 validate the proposed framework.
Keywords: Fokker-Planck optimization, non-linear conjugate gradient,

immunotherapy, chemotherapy
3.1 Introduction
Colon cancer is a leading cause of global cancer related deaths [11]. The lack
of early symptoms forces the detection of colon cancer to take place at the
metastatic phase of the cancer [29]. Thus, it becomes important to devise fast
and accurate treatment strategies for a cure. In this context, combination ther-
apies have been clinically shown to be an effective strategy for combating can-
cer in comparison to monotherapy (see [18]). Conventional monotherapeutic
techniques are indiscriminate in choosing actively growing cells that lead to
the death of not only cancerous cells but also healthy cells. E.g., chemotherapy
drugs can be toxic and lead to multiple side effects and risks by weakening the
*Corresponding author: souvik.roy@uta.edu; ORCID: 0000-0002-6134-1381
33
patient’s immune system, targeting the bone-marrow cells [22]. This leads to
increased susceptibility to secondary infections. However, with combination
therapies, even though it might still be toxic if there is a presence of a chemo-
therapy drug, the toxicity effect is significantly diminished due to different tar-
gets being affected. Moreover, since the effect of combination therapies work
in synergy, lower dosages of the individual drugs in the combination therapy
are required for treatment, which further reduces the toxic effects [19].
To develop optimal combination therapies in colon cancer, it is important
to understand relevant biomarkers that govern the progression of the cancer
[8]. One such biomarker is the immune response to the cancerous cells. It
has been observed that onsite immune reactions at the tumor locations and
prognosis are highly correlated and independent of the size of the tumor
[12]. Furthermore, a high expression of various immune pathways like Th1
and Th17 are associated with a poor prognosis or prolonged disease-free
survival for patients with colon cancer [32]. Thus, it becomes important to
develop optimal personalized treatments for colon cancer patients, taking
into account the various types of immune cells, their numbers, and interac-
tion with the cancer cells and each other. Since drugs for treatment of colon
cancer like chemotherapeutic drugs and immunotherapies are quite costly,
it is expensive to perform in-vitro and vivo experimental studies for testing
effects of such drugs. An alternative cost effective option is to develop com-
putational frameworks for testing optimal combination drug dosages [7].
This can be done using pharmacokinetic models that are given by a set of
ordinary or partial differential equations (ODE or PDE).
There are several dynamic models that have been developed to describe the
immune response and interactions in colon cancer and associated therapies.
In [4], a combined compartmental model with that of irinotecan is used to
describe the physiology of colon cancer. A set of ODEs were used to represent a
3D structure of colon cancer in [14]. In [15], the authors use describe the phe-
nomenon of tumorigenesis in colon cancer using mathematical modeling. The
authors in [17], describe the initiation of colon cancer and its link with colitis
using a mathematical framework. In [23], the authors use a pharmacokinetic
cellular automata model to incorporate the cytotoxic effects of chemotherapy
drugs. In [13], dynamical systems are used to describe multiple pathways in
colon cancer. For a detailed review of various mathematical models for colon
cancer-induced immune response, we refer the reader to [3].
We contribute to the field of pharmacokinetic cancer research by present-
ing an effective approach to develop personalized therapies for colon cancer-
induced immune response. The starting point of this estimation process is the
dynamic model for colon cancer induced immune response, given in [10].
The model describes the evolution of four variables: the tumor cell count,
concentration of the Natural Killer cells, concentration of the CD8+ cells, and
Therapies in Colon Cancer using FP Framework 35
concentration of the remaining lymphocytes. We extend the dynamic model

in [10] to an Itô stochastic ODE system that takes into account the random-
ness of the colon cancer-induced immune response dynamics. We also model
a combination therapy comprising of a chemotherapy drug and an immuno-
therapy drug as a control in the stochastic ODE system. The goal is to obtain
an optimal combination therapy strategy that drives the cancer-induced
immune response state. Since solving an optimal control problem using sto-
chastic state equations is difficult, we use a more convenient framework for
obtaining optimal combination therapies through the Fokker-Planck (FP)
equations that govern the evolution of the joint probability density function
(PDF) associated with the random variables in the stochastic process. Such
FP control frameworks have been used for problems arising in control of
collective and crowd motion [24, 25], investigating pedestrian motion from
a game theoretic perspective [26], reconstructing cell-membrane potentials
and mean field control problems [5], and controlling production of Subtilin
[31]. Very recently, in [27, 28], FP frameworks were used for parameter esti-
mation in colon cancer induced dynamic systems. Until now, this is the first
work that considers the FP framework to devise optimal combination treat-
ment strategies for colon cancer induced immune response.
In the next section, we describe an FP control framework for devising
optimal treatments in colon cancer induced immune response systems.
Section 3.3 is concerned with the theoretical properties of the FP optimiza-
tion problems. In Section 3.4, we describe the numerical discretization of
the optimality system. In Section 3.5, we obtain the optimal combination
therapies comprising of Doxorubicin and IL-2 for two types of simulated
colon cancer patients and show the correspondence of the results with
experimental findings. We end with a section of conclusions.
3.2 The Control Framework Based on

Fokker-Planck Equations
The interaction between immune cells and colon cancer growth in a patient
can be modeled using a coupled system of ODEs. The starting point is the
set of equations given in the paper by dePillis [10]. The following are the
variables associated with different types of cell populations that appear as
model variables.
1. T (τ)- tumor cell population number (cells).

2. N(τ)- natural killer (NK) cells concentration per liter of
blood (cells/L).
3. L (τ)- cytotoxic T lymphocytes (CD8+) concentration per

liter of blood (cells/L).
4. C(τ)- other lymphocytes concentration per liter of blood
(cells/L).
The governing system of ODEs representing the dynamics of the above

defined cell populations is given as follows.
dT
= aT − abT 2 − DT − cNT − α1u1 (t )T , T (0) = T0 ,
dτ
dN
= eC − fN − pNT − α 2u1(t )N + β2u2 (t )N , N (0) = N0 ,
dτ
dL T
= mL + j L − qLT + (r1N + r2C )T − α 3u1(t )L
dτ k +T
+ β3u2 (t )L, L(0) = L0 ,
dC
= α − β C − α 4u1(t )C, C(0) =C0 ,
dτ
(3.1)
1
where D = d ,T0 , N 0 , C0 represent the initial conditions for T, N,
s(T /L )l + 1
L, and C, respectively, and u1, u2 represent dosages of Doxorubicin and
IL-2, respectively.
The parameters of the system (3.1) are defined in [10]. To stabilize the
solutions of the numerical methods, we non-dimensionalize the ODE
system (3.1) using the following non-dimensionalized state variables and
parameters:
=T k=
1T , N k2=
N , L k=
3 L, C k4C , t = k5τ ,
a b c d ek f 1010 p
=a= ,b = ,c = ,d , e = 2= ,f = ,p ,
k5 k1 k2k5 k5 k4 k5 k5 k1k5
m j 108 q r1k3 r2k3
=m = ,j = , k k= 1k , q = , r1 = = , r2 ,
k5 k5 k5k1 k1k2k5 k1k4 k5
α k4 β αi βi
= =
s 250 s, α =,β = l l, α= i = ,. . ., 4, βi
, i 1= i = 1, 2.
k5 k5 k5 k5
(3.2)
Then, the transformed non-dimensional ODE system is given as follows:
dT
= aT − abT 2 − DT − cNT − α1u1(t )T , T (0) = T0
dt
dN
= eC − fN − 10−10 pNT − α 2u1(t )N + β2 u2 (t )N , = N (0) = N0 ,
dt
dL T
= mL + j L − 10−8 qLT + ( r1 N + r2 C)T − α 3 u1(t) L
dt k +T
+ β 3 u 2(t) L , L (0) =L0 ,
dC
= α − β C − α 4 u1 (t )C , C (0) =
C0 , (3.3)
dt
(L / T )ll
where D = d .
4 s ⋅10 ⋅ (k1 / k3 ) + (L / T )
The compact form of the aforementioned system of ODEs, given in
(3.3), is as follows:
dX
= F ( X , U ),
dt
X (0) = X 0 , (3.4)
T
where X (t ) = (T (t ), N (t ), L (t ), C(t )) .
We extend the ODE system (3.3) to include stochasticity present in
the dynamics. For this purpose, we consider the Itô stochastic differential
equation corresponding to (3.3).
dT (aT (1 − bT ) − cNT − DT − α1u1(t )T )dt + σ 1(T )dW1(t ), T (0) = T0 ,

=
dN = (eC − fN − 10−10 pNT − α 2u1(t )N + β2u2 (t )N )dt + σ 2 (N )dW2 (t ),
N (0) = N 0 ,
T
dL = (mL + jT L − 10−8 qLT + (r1N + r2C )T − α 3u1(t )L
k +T
+ β3u2 (t )L)dt + σ 3 (L )dW3 (t ),
L (0) = L0 ,
= (α − β C − α 4u1(t )C )dt + σ 4 (C )dW4 (t ), C (0) = C0 ,
dC
(3.5)
where dWi, i = 1, 2, 3, 4 are one-dimensional Wiener processes and σi, i = 1,

2, 3, 4 are positive constants. Equation (3.5) can be written using a compact
notation as follows:
dX = F (X, U) dt + σ(X) dW (t),
X(0) = X0, (3.6)
where
dW (t) = (dW1(t) dW2(t) dW3(t) dW4(t))T
is a four-dimensional Wiener process vector with stochastically indepen-

dent components and
σ = diag (σ1 σ2 σ3 σ4)
is the dispersion matrix.

We now describe the PDF of the stochastic process (3.6), confined in a
Lipschitz domain Ω by virtue of a reflecting barrier on ∂Ω. This is moti-
vated by the maximum cell carrying capacity.
Thus, X(t) ∈ Ω ⊂ R4+= {x ∈ R4 : xi ≥ 0, i = 1, 2, 3, 4}, Let x = (x1, x2, x3, x4)T.
Define f (x, t) as the PDF for the stochastic process described by (3.6), i.e.,
f (x, t) is the probability of X(t) assuming the value x at time t. Then, the
PDF of X(t) evolves through the following Fokker-Planck (FP) equations.
1
∂ t f (x , t ) + ∇ ⋅ (F (x , U ) f (x , t ))= ∇ ⋅ (σ 2 (x )∇f (x , t )),
2
f (x , 0) = f 0 (x ), (3.7)
where f0(x) is non-negative with mass equals one and U (t ) = (u1 (t ), u2 (t ))

in the admissible set
U ad = {U ∈L2 ([0,T ]): 0 ≤ ui (t ) ≤ Di , Di > 0, ∀t ∈[0,T ], i = 1,2}
Here, the FP domain is Q = Ω × (0, Tf ), where Tf is the final time and

f0(x) represents the distribution of the initial state X0 of the process. The
FP equation (3.7) is associated with the no-flux boundary conditions. To
describe this, we write (3.7) in flux form as
∂tf (x, t) − ∇ · H = 0, f (x, 0) = f0(x), (3.8)
where the components of the flux H is given as follows:
σ 2j
H j (x , t ; f=
) ∂ x j f − Fj (x , U) f , j =
1, 2, 3, 4.
2 (3.9)
Then, the no-flux boundary conditions are
H ⋅n =0 on ∂Ω × (0, T f ),
^
(3.10)
with n̂ as the unit outward normal on ∂Ω.

To obtain the optimal combination therapy function vector U, we solve
the following optimization problem:
α ν T ν T
=U * arg
= minU∈Uad J ( f , U) :
2Q∫( f (x , t ) − f * (x , t ))2 dx + 1 ∫ u1 (t ) 2 dt + 2 ∫ u2 (t ) 2 dt ,
20 2 0
(3.11)
subject to the FP system (3.7),(3.10), where the desired PDF is f *(x, t).
3.3 Theoretical Results

In this section we describe some theoretical results related to the minimiza-
tion problem (3.11). One can also find similar results in [1, 24, 25]. For this
purpose, we denote the FP system (3.7),(3.10) as E(f0, U) = 0. The existence
and uniqueness of solutions of (3.7) is given in the following proposition.
Proposition 3.1. Assume f0 ∈ H1(Ω) with f0 non-negative and U ∈ Uad. Then,

there exists a unique non-negative solution of E(f0, U ) = 0, given by f ∈ L2([0,
Tf ]; H1(Ω)) ∩ C([0, Tf ]; L2(Ω)).
Under the assumptions of higher regularity of ∂Ω and the boundary of
Ω, one can also obtain H2(Ω) regularity of the solution of (3.7) (see [30]).
Next, we state the conservativeness property of (3.7), which can be proved
using straightforward applications of weak formulation, integration by
parts, and divergence theorem for the flux H.
Proposition 3.2. The FP system given in (3.7),(3.10) is conservative.

The next proposition states and proves the L2 stability property of (3.7).
Proposition 3.3. The FP system (3.7),(3.10) solution, given by f, satisfies the

following L2 stability property:
|| f (t )||L2 ( Ω ) ≤ || f0 ||L2 ( Ω ) exp(|| σ −1 ||22 N 2t ),

(3.12)
where N = supΩ×U |F (x, θ)|.

Proof. Multiplying (3.7) with the test function ψ = f (·, t) and an integra-
tion by parts gives:
∂
|| f (t )||2 2 = − ||σ∇f (t )||2 2 + 2Ω (F (t )) ⋅ σ −1σ∇f (t )dx. (3.13)
∂t L ( Ω) L (Ω)
The last term in (3.13) can be estimated using the Young’s inequality,
2bd ≤ kb2 + d2/k, with k = || σ−1 ||2, which is the L2 matrix norm of σ−1. We
then obtain the following:
∂
|| f (t )||22 ≤ ||σ −1 ||2 N 2 || f (t )||22
∂t L (Ω) 2 L (Ω)
Applying the Gronwall’s inequality gives the desired result.

The above results imply that the map Λ : Uad → C([0, Tf ]; H1(Ω)), given
by f = Λ(U), is continuous and Fréchet differentiable. The main theoretical
result in this work, which shows that there is an optimal open-loop control
U*, is given below:
Theorem 3.1. Let f0 ∈ H1(Ω) and let J be given as in (3.11). Then, there exists
(f ∗, U ∗) ∈ C([0, Tf ]; H1(Ω)) × Uad with f ∗ being a solution to E(f0, U ∗) = 0
and U ∗ minimizing J in Uad.
Proof. Since J is bounded, a minimizing sequence (Um) exists in Uad.
Moreover, J is being coercive and sequentially weakens lower semi-continu-
ous in Uad, implying the boundedness of this sequence. Due to the fact that Uad
is a closed and convex subset of a Hilbert space, the sequence (Um) contains a
convergent subsequence (U ml) in Uad, such that U ml → U ∗. Correspondingly,
the sequences (f ml) and (∂t f ml), where fml = Λ(θml) are bounded in L2([0, Tf ];
H1(Ω)), L2([0, Tf ]; H−1(Ω)), respectively. This implies the weak convergence
of the sequences to f ∗ and ∂t f ∗, respectively.
We next use the compactness result of Aubin-Lions [16] to obtain

strongly convergent subsequence (f mk) in L2([0, Tf], L2(Ω)). Thus, the
sequence (F (U mk)fmk) in L2([0, Tf], L2(Ω)) is weakly convergent. This
implies that f ∗ = Λ(U ∗) and (f ∗, U ∗) are minimizers of J.
For the minimization problem (3.11), the optimality system can now be
written as
1
∂ t f ( x , t ) + ∇ ⋅ (F ( x , U ) f (x , t ))= ∇ ⋅ (σ 2∇f (x , t )), in Ω × (0, Tf ),
2
f (x , 0) = f 0 ( x ), in Ω, (FOR )
H ⋅n =
^
0, on ∂Ω × (0, Tf ).
1
−∂t p(x , t ) − f (x , t )(F (x ,θ ) ⋅∇p(x , t )) − ∇ ⋅ (σ 2∇p(x , t ))
2
*
= −α ( f (x , t ) − fi (x , t )), in Ω × (0,T f ),
p(x ,T f ) = 0, in Ω,
∂p
= 0, on ∂Ω × (0,T f ). (ADJ)
∂n
∫
βU − ∇U F ⋅∇p,ψ − U
Ω
L2 ([0,T ])
≥ 0, ∀ψ ∈U ad . (OPT)
The optimality system comprises of three sets of equations: the forward

or state equation that governs the FP dynamics (FOR), the adjoint equa-
tion (ADJ), and the optimality condition (OPT). In the next section, we
describe numerical schemes to implement the optimality system.
3.4 Numerical Schemes

We consider the mesh {Ωh}h>0 given by
Ωh = {(x1, x2, x3, x4) ∈ R4 : (x1i, x2j, x3k, x4l) = (x10 + ih, x20 + jh, x30 + kh,
x40 + lh)},
where (i, j, k, l) ∈ {0, , N x1 } × {0,  , N x2 } × {0, , N x3 } × {0,  , N x4 } ∩ Ω,

and Nxi is the number of discretization points along the ith coordinate
direction. Define δt = Tf /Nt to be the temporal discretization step, where

Nt denotes the maximum number of temporal steps. This gives us the dis-
cretized domain for Ω as follows
Qh,δt = {(x1i, x2j, x3k, x4l, tm) : (x1i, x2j, x3k, x4l) ∈ Ωh, tm = mδt, 0 ≤ m ≤ Nt}.
The value of f (x, t) on Qh,δt is denoted as fim, j . To solve the forward

Fokker-Planck equation (3.7), we use the scheme described in [27], that
is comprised of the spatial discretization using the Chang-Cooper (CC)
method [6]. The temporal derivative discretization is done using the four-
step, alternate direction implicit Douglas-Gunn (DG4) method. The com-
bined scheme is referred to as the DG4-CC scheme. It has been shown in
[27] that this scheme is positive, conservative, stable, and second order
convergent in the L1 norm. Furthermore, we use the temporal D-G scheme
for the time discretization in the first term, one sided finite difference dis-
cretization for the second term, and central difference for the third term on
the left hand side of the adjoint equation (ADJ).
To solve the optimization problem (3.11), we use a projected non-
linear conjugate gradient scheme (PNCG), as described in [28, 27]. Such
a scheme has also been used for solving optimization problems related to
controlling stochastic crowd motion [24, 25], studying avoidance behavior
of pedestrians using game theory [26], and cure rate models [20, 21]. The
PNCG scheme is described below.
Algorithm 3.1 (PNCG Scheme).
1. Input: Initial guess u0. Compute d0 = −∇Jˆ(u0 )H1 . Set k = 0

and maximum number of iterations kmax, tolerance TOL.
2. If (k < kmax), do
3. Compute uk+1 = PU [uk + αk dk] with αk obtained by the Armijo
line-search method.
4. Evaluate g k+1 = ∇Jˆ(uk+1 )H1 .
T
1 
HG || y k ||2 
=
5. Evaluate β y
 k − 2d  g k +1.
dkT y k  dkT y k 
k k
6. Compute dk+1 = − g k+1 β kHG dk .

7. Check for ||uk+1 − uk||2 < TOL. If yes, terminate.
8. Update k = k + 1.
9. End if.
3.5 Results
This section describes the numerical results for obtaining optimal treat-
ment strategies with the aforementioned FP framework. We choose our
domain Ω = (0, 6)4 and discretize it using N xi = 51 points for i = 1, 2, 3, 4.
The final time Tf is chosen to be 10 and the maximum number of time
steps Nt is chosen to be 200. The values of the constants used in converting
the ODE system (3.1) to its non-dimensional form given in (3.3) are given
as k1 = 10−10, k2 = 10−5, k3 = 10−7, k4 = 10−8, and k5 = 1. To obtain the target
PDF f ∗(x, t), we simulate the ODE system (3.3) with the u1 , u2 set to 0 and
with the value of the non-dimensional parameters (d , l , s ) = (2.1,1.1,1.25)
that represents a patient with strong immune response [10]. The values
of the other parameters are taken from [10]. After we obtain the trajec-
tories of T¯, N¯, L¯, and C, we then choose 20 time points ti, and at each
point (T¯(ti), N¯ (ti), L¯(ti), C¯(ti)) assign a Gaussian PDF given by a normal
distribution with variance 0.05. We finally perform a 5D interpolation to
obtain the desired PDF f ∗(x, t). Based on a statistical analysis of the dataset
given in [10], we choose
σ i ( x ) = 0.5( xi1.2 + 0.001), i = 1, 2, 3, 4.
The regularization parameters are chosen to be α = 1 and ν1 = ν2 = 0.01.

The maximum tolerable dosage for Doxorubicin is taken to be 7 mg/day
and for IL-2 is taken to be 7.2×105 IU IL-2/l/day, which implies D1 = 7 and
D2 = 0.072.
In Test Case 1, we simulate a tumor patient with a value of (T (0), N
(0), L(0), C(0)) = (2 × 109, 105, 107, 108). This implies (T¯(0), N¯ (0), L¯(0),
C¯(0)) = (0.2, 1, 1, 1)). We also choose the values of the non-dimensional
parameters (d , l , s ) = (1.3, 2 , 10) that represent patients with weak immune
systems. The goal is to determine optimal drug dosages ui , i = 1,2 such that
the tumor profile is given by the target PDF f ∗.
Figure 3.1 presents the simulation results obtained with our framework.
We observe that without the combination therapy, the mean tumor cell
count will keep on increasing until it reaches the cell carrying capacity,
ultimately leading to patient death. With the combination therapy, the
mean tumor cell count is brought back to diminishing levels. We also note
the optimal dosage patterns for both Doxorubicin and IL-2 over time.
Traditionally, Doxorubicin is administered every 21 days with a dosage of
142.5 mg [9], that translates to a total dosage of 70 mg for 10 days. However,
from Figure 3.1, we note that the total dosage of Doxorubicin over 10 days
×1010 Plot of T (without treatment) ×109 Plot of T (with treatment) Plot of u1 ×105 Plot of u2
4 2 4 2
3.5 1.8 3.5 1.8
1.6 1.6
3 3
1.4 1.4
2.5
u2 (IUIL-2/l)
2.5
u1 (mg)
1.2 1.2
T(cells)
T(cells)
2 1 2
1
1.5 0.8 1.5
0.8
0.6
1 1 0.6
0.4
0.5 0.2 0.5 0.4
0 0 0 0.2
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
t (days) t (days) t (days) t (days)
(a) T (without treatment) (b) T (with treatment) (c) Doxorubicin profile (d) IL-2 profile
Figure 3.1 Test case 1: plots of mean tumor profiles and drug dosages over 10 day period.
×1010 Plot of T (without treatment) ×109 Plot of T (with treatment) Plot of u1 ×104 Plot of u2
3.5 2 3 10
1.8 9
3 2.5
1.6 8
2.5 1.4 7
2
u2 (IUIL-2/l)
1.2 6
u1 (mg)
2
T(cells)
T(cells)
1 1.5 5
1.5 0.8 4
1 3
1 0.6
0.4 0.5 2
0.5 0.2 1
0 0 0 0
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
t (days) t (days) t (days) t (days)
(a) T (without treatment) (b) T (with treatment) (c) Doxorubicin (d) IL-2
Figure 3.2 Test case 2: plots of mean tumor profiles and drug dosages over 10 day period.
is far less than 70 mg. Moreover, the observed daily dosage of IL-2 from
Figure 3.1 is less than the standard dosage of IL-2 over a day, which is 2.1 ×
106 IU IL-2/l. This suggests that optimal combination therapies adminis-
tered daily leads to lower total dosages and, thus, lower toxicity effects.
In Test Case 2, we choose the same initial values of T , N , L , and C, along
with the values of the non-dimensional parameters (d , l , s ) = (1.6 , 1.4 , 2)
that represent patients with moderately strong immune systems.
Figure 3.2 shows the plots of the tumor profiles without and with treat-
ment and the optimal daily dosages of Doxorubicin and IL-2. We observe
a similar behavior as Test Case 1. In addition, we note that the dosages
are smaller compared to the patient in Test Case 1 since the immune sys-
tem is moderately strong. This shows the effectiveness of our framework in
obtaining optimal dosages in colon cancer.
3.6 Conclusion
In this paper, we presented a new stochastic frameworks to determine opti-
mal combination therapies in colon cancer-induced immune response. We
considered the tumor and immune system response dynamics as proposed

in [10] and extend it to a stochastic process to account for incorporating
randomness in the dynamics. We characterized the state of the stochastic
process using the PDF, whose evolution is governed by the FP equation. We
then solved an optimal control problem with open loop controls to obtain
the optimal combination dosages involving chemotherapy and immuno-
therapy. Numerical results demonstrate the feasibility of our proposed
framework to obtain small dosages of the combination drugs, leading to
lower toxicity whilst preserving the effectiveness to eliminate the tumor.
Acknowledgments
The authors were supported by the National Institutes of Health (Grant
Number: R21CA242933).
S. Roy was also partially supported by the Interdisciplinary Research
Program, University of Texas at Arlington, Grant number: 2021-772. The
content is solely the responsibility of the authors and does not necessarily
represent the official views of the National Institutes of Health.
References
1. M. Annunziato and A. Borzì. A fokker–planck control framework for mul-
tidimensional stochastic processes. Journal of Computational and Applied
Mathematics, 237(1):487–507, 2013.
2. M. Annunziato and A. Borzì. A fokker–planck approach to the reconstruc-
tion of a cell membrane potential. SIAM Journal on Scientific Computing,
43(3):B623–B649, 2021.
3. A. Ballesta and J. Clairambault. Physiologically based mathematical mod-
els to optimize therapies against metastatic colorectal cancer: a mini-review.
Current pharmaceutical design, 20(1):37–48, 2014.
4. A. Ballesta, J. Clairambault, S. Dulong, and F. Levi. A systems biomedicine
approach for chronotherapeutics optimization: focus on the anticancer drug
irinotecan. In New Challenges for Cancer Systems Biomedicine, pages 301–
327. Springer, 2012.
5. A. Borzì and L. Grüne. Towards a solution of mean-field control problems
using model predictive control. IFAC-PapersOnLine, 53(2):4973–4978, 2020.
6. J. Chang and G. Cooper. A practical difference scheme for fokker-planck
equations. Journal of Computational Physics, 6(1):1–16, 1970.
7. Y. Chang, M. Funk, S. Roy, E. Stephenson, S. Choi, H. V. Kojouharov, B.
Chen, and Z. Pan. Developing a mathematical model of intracellular calcium
dynamics for evaluating combined anticancer effects of afatinib and rp4010

in esophageal cancer. International Journal of Molecular Sciences, 23(3):1763,
2022.
8. Y. Chang, S. Roy, and Z. Pan. Store-operated calcium channels as drug target
in gastroesophageal cancers. Frontiers in pharmacology, 12:944, 2021.
9. L. de Pillis, K. Renee Fister, W. Gu, C. Collins, M. Daub, D. Gross, J. Moore,
and B. Preskill. Mathematical model creation for cancer chemo-immuno-
therapy. Computational and Mathematical Methods in Medicine, 10(3):165–
184, 2009.
10. L. DePillis, H. Savage, and A. Radunskaya. Mathematical model of colorectal
cancer with monoclonal antibody treatments. arXiv preprint arXiv:1312.3023,
2013.
11. C. Fitzmaurice, D. Dicker, A. Pain, H. Hamavid, M. Moradi-Lakeh, M. F.
MacIntyre, C. Allen, G. Hansen, R. Woodbrook, C. Wolfe, et al. The global
burden of cancer 2013. JAMA oncology, 1(4):505–527, 2015.
12. J. Galon, A. Costes, F. Sanchez-Cabo, A. Kirilovsky, B. Mlecnik, C. Lagorce-
Pagès, M. Tosolini, M. Camus, A. Berger, P. Wind, et al. Type, density, and
location of immune cells within human colorectal tumors predict clinical
outcome. Science, 313(5795):1960–1964, 2006.
13. S. Haupt, A. Zeilmann, A. Ahadova, H. Bläker, M. von Knebel Doeberitz,
M. Kloor, and V. Heuveline. Mathematical modeling of multiple pathways in
colorectal carcinogenesis using dynamical systems with kronecker structure.
PLoS computational biology, 17(5):e1008970, 2021.
14. M. D. Johnston, C. M. Edwards, W. F. Bodmer, P. K. Maini, and S. J. Chapman.
Mathematical modeling of cell population dynamics in the colonic crypt
and in colorectal cancer. Proceedings of the National Academy of Sciences,
104(10):4008–4013, 2007.
15. N. L. Komarova, C. Lengauer, B. Vogelstein, and M. A. Nowak. Dynamics of
genetic instability in sporadic and familial colorectal cancer. Cancer biology
& therapy, 1(6):685– 692, 2002.
16. J.-L. Lions. Quelques méthodes de résolution de problemes aux limites non
linéaires. Paris, Dunod-Gauth. Vill., 1969.
17. W.-C. Lo, E. W. Martin Jr, C. L. Hitchcock, and A. Friedman. Mathematical
model of colitis-associated colon cancer. Journal of theoretical biology,
317:20–29, 2013.
18. R. B. Mokhtari, T. S. Homayouni, N. Baluch, E. Morgatskaya, S. Kumar, B.
Das, and H. Yeger. Combination therapy in combating cancer. Oncotarget,
8(23):38022, 2017.
19. R. B. Mokhtari, S. Kumar, S. S. Islam, M. Yazdanpanah, K. Adeli, E. Cutz, and
H. Yeger. Combination of carbonic anhydrase inhibitor, acetazolamide, and
sulforaphane, reduces the viability and growth of bronchial carcinoid cell
lines. BMC cancer, 13(1):1–18, 2013.
20. S. Pal and S. Roy. A new non-linear conjugate gradient algorithm for
destructive cure rate model and a simulation study: illustration with nega-
tive binomial competing risks. Communications in Statistics-Simulation and
Computation, pages 1–15, 2020.
21. S. Pal and S. Roy. On the estimation of destructive cure rate model: A
new study with exponentially weighted poisson competing risks. Statistica
Neerlandica, 75(3):324–342, 2021.
22. A. H. Partridge, H. J. Burstein, and E. P. Winer. Side effects of chemotherapy
and combined chemohormonal therapy in women with early-stage breast
cancer. JNCI Monographs, 2001(30):135–142, 2001.
23. G. G. Powathil, D. J. Adamson, and M. A. Chaplain. Towards predicting the
response of a solid tumour to chemotherapy and radiotherapy treatments:
clinical insights from a computational model. PLoS computational biology,
9(7):e1003120, 2013.
24. S. Roy, M. Annunziato, and A. Borzì. A fokker–planck feedback control-
constrained approach for modelling crowd motion. Journal of Computational
and Theoretical Transport, 45(6):442–458, 2016.
25. S. Roy, M. Annunziato, A. Borzì, and C. Klingenberg. A fokker–planck
approach to control collective motion. Computational Optimization and
Applications, 69(2):423–459, 2018.
26. S. Roy, A. Borzì, and A. Habbal. Pedestrian motion constrained by fp-
constrained nash games. Royal Society Open Science, 4:170648, 2017.
27. S. Roy, S. Pal, A. Manoj, S. Kakarla, J. Villegas, and M. Alajmi. A fokker-planck
framework for parameter estimation and sensitivity analysis in colon cancer.
AIP Conference Proceedings (accepted, in press), 2522, 070005, 2022.
28. S. Roy, Z. Pan, and S. Pal. A fokker-planck feedback control framework
for optimal personalized therapies in colon cancer-induced angiogenesis.
Journal of Mathematical Biology, 84:23, 2022.
29. H. Schmoll, E. Van Cutsem, A. Stein, V. Valentini, B. Glimelius, K.
Haustermans, B. Nordlinger, C. Van de Velde, J. Balmana, J. Regula, et al.
Esmo consensus guidelines for management of patients with colon and rectal
cancer. a personalized approach to clinical decision making. Annals of oncol-
ogy, 23(10):2479–2516, 2012.
30. T. Tao. Nonlinear dispersive equations: local and global analysis. Number 106.
American Mathematical Soc., 2006.
31. V. Thalhofer, M. Annunziato, and A. Borzì. Stochastic modelling and control
of antibiotic subtilin production. Journal of mathematical biology, 73(3):727–
749, 2016.
32. M. Tosolini, A. Kirilovsky, B. Mlecnik, T. Fredriksen, S. Mauger, G. Bindea,
A. Berger, P. Bruneval, W.-H. Fridman, F. Pagès, et al. Clinical impact of dif-
ferent classes of infiltrating t cytotoxic and helper cells (th1, th2, treg, th17)
in patients with colorectal cancer. Cancer research, 71(4):1263–1271, 2011.
4
Detection and Classification of
Leaf Blast Disease using Decision
Tree Algorithm in Rice Crop
Sarvesh Vishwakarma1* and Bhavna Chilwal2
Department of CSE, Graphic Era (Deemed to be University), Dehradun, India

1
2
Department of CSE, DIT University Dehradun, Uttarakhand, India
Abstract
The agricultural field is the most important field for any nation but some issues pre-
vail and affect agricultural products every year. Agricultural diseases are the main
concern for yield loss. This chapter uses the Decision Tree technique to form a tree
structure for leaf blast disease level detection in rice crops. A Decision Tree is used
as a classification technique and here disease levels are classified based on symp-
toms that occur during infection. The Iterative Dichotomiser 3 (ID3) algorithm
is one of the important methods to form a Decision Tree based on entropy and
information gain. The nodes in the tree are the symptoms that have different labels
for disease occurrence. This decision tree will help detect the occurrence of disease
as per the symptoms and help farmers get information about the severity level of a
disease so that they can take required measures on time to save the crop from loss.
Keywords: Decision tree, ID3 algorithm, entropy, information gain, leaf blast
disease
4.1 Introduction
There are so many fungal diseases that prevail in rice, but this chapter
focuses on leaf blast disease in rice crops. Blast disease is also known as
rice fever. Agricultural scientists find out about the disease by checking
their symptoms, similarly, this paper uses three particular symptoms for
leaf blast [5]. These symptoms will be used to form a decision tree and
*Corresponding author: sarveshvishwakarma.cse@geu.ac.in

49
(a) Healthy Leaf (b) Disease Leaf
Figure 4.1 Comparison between healthy and diseased leaf.
detect the labels of disease occurrence (as shown in Figure 4.1(a) and
Figure 4.1(b)).
The decision tree uses attributes to classify the leaves according to
symptoms. Every decision tree comprises of two types of nodes: a Decision
Node and the Decision Leaf Node. Different classification algorithms are
used nowadays to implement decision trees. Here, an ID3 classification
algorithm is used for implementation.
This algorithm stands for Iterative Dichotomiser 3. This algorithm forms
the smallest decision tree possible. The ID3 algorithm uses the Entropy and
Information Gain [6–8, 13]. The role of Entropy is to control the splitting
of the data for the decision tree. With the help of entropy, the boundaries of
the decision tree are decided. The formula for class entropy is given below:
−P  P  N  N 
T= log 2  − log 2   (4.1)
P+N P+N  P+N P+N 
For a sample of negative N- and positive P+ Entropy for each attribute:
−Pi  Pi  Ni  Ni 
IG (Pi , N i ) = log 2   − P + N log 2  P + N  (4.2)
Pi + N i  Pi + N i  i i  i i 
E(S) =
∑ (P + N ) × I (P , N )
i i
G i i (4.3)
P+N
The information gain calculates the decrease in the value of entropy

after data gets split on different attribute values. The decision tree finds the
Detection and Classification of Leaf Blast Disease 51
parameter value which has the highest information gain. The three symp-
toms taken for disease detection are the growth stage of the rice crop, disease
index, and lesion type [1]. Linguistic variables are taken for different ranges
of these variables. The ID3 algorithm will detect the leaf blast disease in rice
crops by using these symptoms as the attributes and decide from which
class the rice leaf belongs. The implementation of a decision tree in different
research is very vast. It framed different rules to form a decision tree that is
influenced by climatic parameters for the productivity of soybeans [2]. The
tree takes symptoms and forms forty rules for fuzzy logic for detecting leaf
blast disease by using regression methodology [3]. A decision tree classifier
is used by measuring the temperature and soil moisture as parameters to
form a system for predicting cotton crop disease [4]. Medical data mining
for the prediction of heart disease is done by using Naïve Bayes and decision
tree algorithms [8, 11, 12] and this research work shows a prediction accu-
racy of 99%. A prototype was used to evaluate the severity of disease on rice
crops by using computational intelligence and machine learning. K-Mean
segmentation has been used with fuzzy logic to calculate the degree of the
disease that occurred in rice plants [9, 10]. The prototype has 86% accuracy.
4.2 Proposed Methodology

The symptoms and rule table are considered based on the work described
in [14, 15], which presents fuzzy rule techniques for different combina-
tions of symptoms [16–18]. First of all, we take the attributes of a dataset
and form a table that comprises linguistic values regarding each attribute
and response column with positive or negative responses [19, 20]. Then,
we calculate the class entropy for the whole table, i.e., Entropy (T). But
among all the attributes we have to find the root attribute for the tree, we
calculate the information gain for each attribute which is then used to find
the entropy value of that specific attribute, i.e., Entropy (X). The computa-
tion of gain has to be performed after getting the attributed entropy values
and comparing them against the class entropy (as shown in Equation 4.4).
GainAttribute = T − E (S) (4.4)
The attribute which has maximum gain value becomes the root node
and again the iteration of the whole process starts for expanding the tree
[21–23]. So, to form the table we have provided linguistic terms for differ-
ent values of each attribute or symptom. The rice plant has three important
growth stages and during these stages, the plant has some height between
1 cm to 100 cm on average. Therefore, the linguistic variables [24–26] for

the Growth Stage are mentioned below:
• Germinating Stage – Small (S)

• Vegetative Stage – Moderate (M)
• Reproduction Stage – Tall (T)
For the disease index, the range of input scores is varying from 0 to
100%. Scores are predefined in the standard evaluation system by IRRI
2015. The linguistic variables for different scales are provided below:
• Score 1 and 3 – Low

• Score 5 and 7 – Mid
• Score 9 – High
Lesion type (standard evaluation system of rice (SES), IRRI 2015) values
are taken from scale 5 to scale 9 because as per the experts, the lesions are
noticeable properly.
• Scale 5 and 6 – Low

• Scale 7 and 8 – Mid
• Scale 9 – High
The result column has two specific classes for which two values are used
to provide disease labels:
1. Resistive R (+) for healthy leaves or leaves with a very small

effect
2. Susceptive S (-) for diseased leaves which need proper care
and are harmful for crop production
4.3 Result Analysis

After deciding the linguistic values for different symptoms, we formed a
rule base table (as shown in Table 4.1), which has fourteen combinations of
these symptoms with their disease occurrence labels. The computation of
class entropy for the disease occurrence has been done as below:
Resistive (+) = 4
Susceptive (-) = 10
Table 4.1 Rule-based disease occurrence.

Growth stage Disease index Lesion type Occurrence label
Small No No R (+)
Moderate Low No R (+)
Moderate Low Low R (+)
Tall Mid Mid S (-)
Moderate Low Low S (-)
Tall Mid Mid S (-)
Moderate Mid Mid S (-)
Tall Mid Mid S (-)
Tall High High S (-)
Tall Mid Mid S (-)
Moderate Low Mid S (-)
Moderate Low Low R (+)
Moderate High Low S (-)
E (Disease Occurrence) = E (4, 10)

E (4, 10) = − (0.714 log2 0.714) − (0.286 log2 0.286) = 0.863
This value is used to find the Gain for each attribute separately, as shown
in Table 4.2.
The choice of parameter is done to reflect the highest gain value for the
decision tree. Therefore, the ID3 algorithm is applied for creating a deci-
sion tree and we detect the symptoms which have maximum impact and
Table 4.2 Gain value for each attribute.

Growth stage
Growth stage R(+) S(-) IG Entropy Gain
Low 0 0 0 0.496 0.367
Moderate 4 3 0.951
Tall 0 7 0
that will become the root of the tree, i.e., lesions type as it has the highest
information gain value. The attribute which has the greatest gain as the
decision node becomes the root of the decision tree so the tree would look
like a three-child leaf (Low, Mid, and High) with root node Lesions Type.
From Table 4.3, we get all mid and high-range lesion types having dis-
ease occurrence response as the susceptive (+) class label response, mean-
ing the farmers have to take measures to stop the severity of the disease.
But the low range of lesions type has two responses, both resistive and sus-
ceptive labels, for some rows so we take the next attribute branch which
has the second greatest information gain, i.e., disease index which is 0.46,
as depicted in Table 4.4.
Table 4.3 Lesion types for each attribute.

Lesion type
Lesion type R(+) S(-) IG Entropy Gain
Low 4 1 0.721 0.281 0.582
Mid 0 7 0
High 0 2 0
Table 4.4 Disease index for each attribute.

Disease type
Disease index R(+) S(-) IG Entropy Gain
Low 4 2 0.9198 0.397 0.466
Mid 0 5 0
High 0 3 0
Lesion Type
Low Mid High
Disease Index Susceptive Susceptive
Low Mid High
Figure 4.2 Detection of root node for decision tree.

Lesion Type
Low Mid High

Disease Index Susceptive Susceptive
Low Mid High

Growth Stage Susceptive Susceptive
Small Moderate Tall

No Infection Susceptive
Resistive Susceptive
Figure 4.3 Final decision tree for leaf blast disease.
Figure 4.2 has been derived by using two major symptoms of leaf blast
disease in which lesion type is the root node.
Now, in the next iteration we take the third greatest information gain
attribute branch, i.e., the growth stage with a gain value of 0.367. The tree
would take the shape, as shown in Figure 4.3.
4.4 Conclusion
Disease in crops and plants is a major problem faced by farmers. The tra-
ditional methods to detect the occurrence of diseases are tiresome, costly,
and are not time efficient in comparison with the new soft computing tech-
niques which provide a digital era in the agricultural sector. The decision
tree provides an easier view of the whole process and how to detect the
occurrence level of disease. We can say that the crop has specific symptoms
during a particular type of disease. This means by making use of symp-
toms, the risk level could be detected which is very helpful for farmers and
agricultural scientists to save the yield loss.
4.5 Future Work

The future scope for this work is to implement the tree by using dif-
ferent algorithms, except for ID3. Also, we can form the decision tree
for the classification of other diseases. We can also extend this work by
using a Random Forest algorithm by making more than one decision
tree.
References
1. Pari Skamnioti, Sarah J. Gurr, Against the grain: safeguarding rice from rice
blast disease, Trends in Biotechnology, Volume 27, Issue 3, 2009, Pages 141-
150, ISSN 0167-7799, https://doi.org/10.1016/j.tibtech.2008.12.002.
2. P. Udupa, S. Vishwakarma, A survey of MRI segmentation Techniques for
Brain Tumor Studies, Bonfring International Journal of Advances in Image
Processing, Vol. 6, No. 3, August 2016, pp. 22-27. Doi: http://www.journal.
bonfring.org/papers/aip/volume6/BIJ-10467.pdf
3. F. M. Shamim, S. Vishwakarma, Exploiting the Motion Learning Paradigm
for Recognizing Human Actions, Bonfring Internation Journal of Advances
in Image Processing, Vol. 6, No. 3, August 2016, pp. 11-16. http://www.journal.
bonfring.org/papers/aip/volume6/BIJ-10465.pdf
4. Chopda, J., Raveshiya, H., Nakum, S., Nakrani, V., 2018. Cotton crop disease
detection using decision tree classifier. International Conference on Smart
City and Emerging Technology (ICSCET), 1–6.
5. Donatelli, M., Magarey, R.D., Bregaglio, S., L., W., Whish, J.P.M., Savary, S.,
2017. Modelling the impacts of pests and diseases on agricultural systems.
Agricultural Systems 155, 213–224.
6. Edwards-Murphy, F., Magno, M., Whelan, P.M., O’Halloran, J., Popovici,
E.M., 2016. b+wsn: Smart beehive with preliminary decision tree analysis for
agriculture and honey bee health monitoring. Computers and Electronics in
Agriculture 124, 211–219.
7. Ilic, M., Ilic, S., Jovic, S., Panic, S., 2018. Early cherry fruit pathogen disease
detection based on data mining prediction. Computers and Electronics in
Agriculture (Elsevier) 150, 418–425.
8. Karthiga, A.S., Mary, M.S., Yogasini, M., 2011. Early prediction of heart
disease using decision tree algorithm. International Journal of Advanced
Research in Basic Engineering Sciences and Technology (IJARBEST) 3(03).
9. Nithya, A., Sundaram, V., 2011. Wheat disease identification using classifica-
tion rules. International Journal of Scientific & Engineering Research 2(09),
1–5.
10. Sethy, P.K., Negi, B.S., 2018. Measurement of disease severity of rice crop
using machine learning and computational intelligence. Cognitive Science
and Artificial Intelligence (Springer) 2, 1–11.
11. Shouman, M., Turner, T., Stocker, R., 2011. Using decision tree for diag-
nosing heart disease patients. 9th Australasian Data Mining Conference
(AusDM 2011) 121, 23–29.
12. Singh, P., Jagyasi, B., Rai, N., Gharge, S., 1–6. Decision tree-based mobile
crowdsourcing for agriculture advisory system. 2014 Annual IEEE India
Conference (INDICON), 2014.
13. Zhang, Y.D., Wang, S., Dong, Z., 2014. Classification of alzheimer’s disease
based on structural magnetic resonance imaging by kernel support vector
machine decision tree. Progress In Electromagnetics Research 144, 171–184.
14. Vishwakarma, S., Agrawal, A. A survey on activity recognition and behav-

ior understanding in video surveillance. Vis Comput 29, 983–1009 (2013).
https://doi.org/10.1007/s00371-012-0752-6
15. S. Vishwakarma and A. Agrawal, A novel approach for feature quantiza-
tion using one-dimensional histogram, 2011 Annual IEEE India Conference,
2011, pp. 1-4, doi: 10.1109/INDCON.2011.6139391.
16. S. Vishwakarma, A. Sapre and A. Agrawal, Action recognition using cuboids
of interest points, 2011 IEEE International Conference on Signal Processing,
Communications and Computing (ICSPCC), 2011, pp. 1-6, doi: 10.1109/
ICSPCC.2011.6061680.
17. V. Shetty, S. Vishwakarma, Harisha and A. Agrawal, Design and imple-
mentation of video synopsis using online video inpainting, 2017 2nd IEEE
International Conference on Recent Trends in Electronics, Information &
Communication Technology (RTEICT), 2017, pp. 1208-1212, doi: 10.1109/
RTEICT.2017.8256790.
18. S. Vishwakarma and A. Agrawal, Framework for human action recognition
using spatial temporal based cuboids, 2011 International Conference on Image
Information Processing, 2011, pp. 1-6, doi: 10.1109/ICIIP.2011.6108881.
19. A. Genovese, V. Piuri, F. Scotti and S. Vishwakarma, Touchless Palmprint
and Finger Texture Recognition: A Deep Learning Fusion Approach, 2019
IEEE International Conference on Computational Intelligence and Virtual
Environments for Measurement Systems and Applications (CIVEMSA),
2019, pp. 1-6, doi: 10.1109/CIVEMSA45640.2019.9071620.
20. R. Singh, S. Vishwakarma, A. Agrawal, and M.D. Tiwari, Unusual activity detec-
tion for video surveillance, IITM ‘10: Proceedings of the First International
Conference on Intelligent Interactive Technologies and Multimedia,
December 2010 Pages 297–305 https://doi.org/10.1145/1963564.1963616
21. Donida Labati R., Genovese A., Piuri V., Scotti F., Vishwakarma S. (2020)
Computational Intelligence in Cloud Computing. In: Kovács L., Haidegger
T., Szakál A. (eds) Recent Advances in Intelligent Engineering. Topics in
Intelligent Engineering and Informatics, vol 14. Springer, Cham. https://
doi.org/10.1007/978-3-030-14350-3_6
22. Vishwakarma, S. and Agrawal, A., 2010. A Robust Algorithm for
Shadow Removal of Moving Object Detection in Video Surveillance
System. International Journal of Recent Trends in Engineering and
Technology, ACEEE, AMAE, and ACEE, USA, 4(1), pp.73-75.
23. Donida Labati, Angelo Genovese, Vincenzo Piuri, Fabio Scotti, Sarvesh
Vishwakarma, I-SOCIAL-DB: A labeled database of images collected
from websites and social media for Iris recognition, Image and Vision
Computing, Volume 105, 2021, 104058, ISSN 0262-8856, https://doi.
org/10.1016/j.imavis.2020.104058
24. Vishwakarma S., Agrawal A. (2012) Representing Feature Quantization
Approach Using Spatial-Temporal Relation for Action Recognition.
In: Kundu M.K., Mitra S., Mazumdar D., Pal S.K. (eds) Perception
and Machine Intelligence. PerMIn 2012. Lecture Notes in Computer

Science, vol 7143. Springer, Berlin, Heidelberg. https://doi.org/10.1007/
978-3-642-27387-2_13
25. Acharya P., Vishwakarma S., Abnormal Crowd Behavior Detection using
Structural Context Descriptor, Bonfring Internation Journal of Advances
in Image Processing, Vol. 6, No. 3, August 2016, pp. 17-21 DOI: 10.9756/
BIJAIP.10466
26. Udupa P., Vishwakarma S., Segmentation of Brain MRI to differentiate
Healthy and Tumor Affected Tissues along with Subsequent Tumor Grading,
International Journal of Image Processing and Applications, Vol. 1, Issue 01,
October 2016, pp. 1-6.
5
Novel Hybrid Optimal Deep Network
and Optimization Approach for
Human Face Emotion Recognition
J. Seetha1, M. Ayyadurai2* and M. Mary Victoria Florence3
Department of Computer Science and Business Systems,

1
Panimalar Engineering College, Chennai, India

2
Department of CSE, SRM Institute of Science and Technology, Ramapuram,
Chennai, India
3
Department of Mathematics, Panimalar Engineering College, Chennai, India
Abstract
Emotion is significant in deciding a human’s ideas, behavior, and feelings. Using
the advantages of deep learning, an emotion detection system can be constructed
and various applications such as face unlocking, feedback analysis, and so on are
executed with high accuracy. Artificial intelligence’s fast development has made a
significant contribution to the technological world. However, it has several diffi-
culties in achieving optimal recognition. Interpersonal differences, the intricacy
of facial emotions, posture, and lighting, among other factors, provide signifi-
cant obstacles. To resolve these issues, a novel Hybrid Deep Convolutional based
Golden Eagle Network (HDC-GEN) model algorithm is proposed for the effective
recognition of human emotions. The main goal of this research is to create hybrid
optimal strategies that classify five diverse human facial reactions. The feature
extraction of this research is carried out using Heap Coupled Bat Optimization
(HBO) method. The execution of this research is performed by MATLAB soft-
ware. The simulation outcomes are compared with the conventional methods in
terms of accuracy, recall, precision, and F-measure and the comparison shows the
effective performance of proposed approaches in facial emotion recognition.
Keywords: Face emotion recognition, deep learning, optimization, hybrid model,

golden eagle optimization, feature extraction, and bat optimization
*Corresponding author: ayyadurm@srmist.edu.in
59
5.1 Introduction
Emotional analytics combine psychology and technology in an intriguing
way. The examination of facial gestures is one of the most used methods
for recognition of emotions. Basic emotions are thought to be physiolog-
ically fixed, intrinsic, and universal to all individuals and many animals
[1]. Severe reactions are either a collection of fundamental emotions or a
collection of unusual feelings. The major issue is figuring out which emo-
tions are fundamental and which are complicated. While analyzing the
information acquired by the ears and sight, people may detect these signals
even when they are softly expressed. Based on psychological research that
illustrates that visual data affects speech intelligibility, it is reasonable to
infer that human emotion interpretation shares a predictable pattern [2].
The practice of recognizing human emotions from facial gestures is known
as recognition of facial expression. Facial expression is a universal signal
that all humans use to communicate their mood. In this era, facial expres-
sion detection systems are very important because they can record people’s
behavior, sentiments, and intentions [3].
Computer networks, software, and networking are rapidly evolving and
becoming more widely used. These systems play a vital part in our daily lives
and make life much easier for us. As a result, face expression detection as a
method of image analysis is quickly expanding. Human–computer interface,
psychological assessments, driving enabled businesses, automation, drunk
driver identification, healthcare, and, most importantly, lie detection are
all conceivable uses [4]. The human brain perceives emotions instinctively
and technology that can recognize emotions has recently been developed.
Furthermore, Artificial Intelligence (AI) can identify emotions by under-
standing the meaning of each face expression and learn to adapt to fresh input.
Artificial Neural Networks (ANN) are being employed in AI systems
currently [5]. Long Short Term Memory (LSTM) based RNN [6], Nave
Bayes, K-Nearest Neighbors (KNN), and Convolutional Neural Networks
[7] are used to tackle a variety of challenges such as excessive makeup,
position, and expression [8]. The properties of the swarm intelligence opti-
mization method, which can execute parallel computation and sharing of
information, have caught our interest in order to make good use of excess
computational resources and enhance optimization effectiveness [9]. The
Particle Swarm Optimization (PSO) and Ant Colony Optimization (ACO)
algorithms, as well as their upgraded variants, constitute the swarm intelli-
gence algorithm [10]. The traditional approaches are slow and inaccurate,
but a face expression detection system based on deep learning has proven
Human Face Emotion Recognition 61
to be superior. Furthermore, typical techniques largely focus on face analy-

sis while keeping the backdrop intact, resulting in a large number of irrele-
vant and inaccurate characteristics that complicate CNN training.
In this framework, this paper proposes a novel for facial emotion recog-
nition using a Hybrid Deep Convolutional based Golden Eagle Network
(HDC-GEN) model algorithm. Moreover, the finest features are extracted
using HBO method. The current article comprises of four basic facial expres-
sion categories that have been reported: displeasure/anger, sadness/unhap-
piness, happiness/smile, fear, and surprise/astonishment. Investigations have
been carried out and the findings reveal that the suggested network outper-
forms the benchmark and state-of-the-art solutions for emotion recognition
based on facial expression images. It offers several advantages over manual
correction in terms of training efficiency and performance, especially when
computational resources are sufficient. When compared to other standard
emotion detection approaches, this proposed method produced better results
in analyzing and classifying emotional data gathered from investigations.
The following is an outline of the article’s layout: The traditional meth-
ods used for emotion classification are explained in Section 5.2. In Section
5.3, the system model along with problem definition is provided. The pro-
posed method is described in Section 5.4. Finally, the findings are provided
in Section 5.5 and the implications are explored in Section 5.6.
5.2 Related Work

Some of the recent works correlated to this research are articulated as
follows: Humans have traditionally had an easy time detecting emotions
from facial expression but performing the same performance with a com-
puter program is rather difficult. In this study, Mehendale and Ninad [11]
presented a unique face emotion identification approach based on con-
volutional neural networks. Using an EV of length 24 data, it was feasi-
ble to appropriately emphasize the emotion with 96 percent accuracy.
Furthermore, before to the formation of EV, a unique background removal
approach is used to prevent dealing with many issues that may arise.
Researchers were able to attain cutting-edge performance in emotion
identification by employing deeper architectures. For this reason, Mungra,
Dhara, et al. [12] developed the PRATIT model which employs certain
image preprocessing processes and a CNN design for face emotion iden-
tification. To manage differences in the photos, preprocessing methods
including gray scaling, resizing, cropping, and normalization are utilized
in PRATIT, which obtains an accuracy rate of 78.52%.
A multimodal attention based Bidirectional Long Short-Term Memory

(BLSTM) network architecture for effective emotion identification is
described by Li, Chao, et al. [13]. The optimal temporal properties are auto-
matically learned using BLSTM with Recurrent Neural Networks (LSTM-
NNs). The learnt deep characteristics are then given into a DNN, which
predicts the probability of emotional response for every stream. Furthermore,
the overall feeling is predicted using a decision level hybrid method.
Estrada, et al. [14] compared numerous sentiment classification mod-
els for the categorization of learning attitudes in an Artificial Learning
Environment termed ILE-Java, employing three distinct methodologies
including deep learning, machine learning, and an evolutionary technique
named EvoMSA. It was also found that the EvoMSA produced the best find-
ings, with an accuracy of 93% for the sample sentiTEXT and 84% for the
sample eduSERE, depending on the outcomes of the mentioned methods.
In reality, optimizing hyper-parameters continues to be a problem when
developing machine learning models like CNNs. Therefore, Gao, Zhongke, et
al. [15] offered an autonomously optimized model that selects the structure
utilizing a binary coding scheme and GPSO with gradient consequences. The
GPSO-based algorithm effectively explores the optimum solution, enabling
CNNs to outperform well in classification throughout the dataset. The find-
ings demonstrate that this strategy, which is centered on a GPSO-optimized
Classification algorithm, reached a high level of recognition rate.
5.3 System Model and Problem Statement

Humans accomplish the task of identifying expressions on a daily basis
with ease, yet automated expression identification remains tough due to
the difficulties in separating the emotions’ feature set. Face detection and
preprocessing, face image extraction, and classification techniques are the
three phases involved in human recognition utilizing facial images. There
are major differences in how people articulate themselves. Due to numer-
ous elements such as backdrop, lighting, and position, photographs of
the same subject with the same expression change. Emotion recognition
is difficult owing to a variety of input modalities that play a vital part in
interpreting information. The task of recognizing emotions is complicated
by the lack of a big collection of training images and the difficulty of dis-
tinguishing emotion depending on whether the input image is fixed or an
evolving clip into face features. The challenge faced is mostly for real-time
identification of facial emotions that vary drastically.
5.4 Proposed Model

Emotion artificial intelligence is a system that can detect, imitate, under-
stand, and deal with human facial gestures and feelings. Artificial intel-
ligence systems have grown in popularity in the contemporary world as
a result of their extensive capabilities and ease of use. However, the tra-
ditional methods have certain drawbacks for the accurate recognition of
emotions. Thus, a novel HDC-GEN classification method is proposed in
this work for significant human facial emotion identification. Primarily,
the input images are preprocessed using a median filter because the raw
dataset has more noise and unwanted elements are placed. Furthermore,
the knowledge based face detection function and image cropping are pro-
vided using a MATLAB function. The feature extraction is employed by
HBO algorithm to improve the classification. Then, the HDC-GEN algo-
rithm improves the facial expression recognition accuracy. The proposed
model of emotion recognition diagrammatic illustration is provided in
Figure 5.1.
Human emotion
dataset
Input image
Pre-processing
stage
by median filter
Feature
extraction by
Resize image
HBO method Face detection // knowledge
based algorithm
Performance
HDC-GEN method Emotion
measures
Parameter tuning
Figure 5.1 Proposed model of emotion recognition illustration.

5.4.1 Preprocessing Stage

Preprocessing is the initial step in the process of extracting features from
images. Moreover, preprocessing is the phase at which noise is removed
via filtering. The median filter is employed to remove noise in this case.
Preprocessing is necessary because it removes unnecessary noise and
improves the sharpness of the images.
5.4.2 Knowledge Based Face Detection

The aim of face detection in this research is to give the position and size
of the face in the entire image. Learning-based face detection techniques
are the most successful methods in terms of detection accuracy and speed.
There could be a variety of items in the background of the input image,
such as a building, a people, or trees. This face detection framework can
process images very quickly while attaining good detection rates. In this
study, the goal of face detection is to determine the size and position
of the face in the full image. In terms of detection speed and accuracy,
knowledge-based image detection algorithms are the most successful.
Rectangular characteristics can be generated quickly using this cascade
object detector in MATLAB. Any position’s image representation is the
summation of the pixels beyond and to the side of it. For example, the sum
of the pixels beyond and to the side of m and n makes up the integral image
at point m and n.
5.4.3 Image Resizing

The cropping function is used to crop the face area from the input human
image once the bounding box has detected the face region using MATLAB.
The facial region inside each sample image is clipped at this point.
5.4.4 Feature Extraction

A right and left eye, nose, and mouth are among the traits examined in the
suggested technique. The images are then enlarged to maintain the same
distance between the midpoints of the left and right eyes and rotated such
that the line connecting the two is horizontal. Next, we initialize all feature
parameters to the HBO method. The feature extraction is done by the max-
imum and minimum point estimation using Equation 5.1:
di = dmin (dmax − dmin) × δ (5.1)

where di is the feature point, minimum distance rate is denoted as dmin,

maximum distance rate is denoted as dmax, and random path of selection is
considered as δ. Furthermore, the priority of feature extraction estimated
by the fitness in Equations 5.2 and 5.3:
f1n  x1n −1 − X∗  × fi + f1n −1NT

= (5.2)
x1n x1n −1 + f1n

= (5.3)
where n is denoted as time step, f1n is the optimal fitness speed, and the
best point is considered as X*. The extraction condition for each feature is
evaluated by the fitness of heap using Equation 5.4:
^
Mni n
Mni fin (t) , f Mni f f in (t )
n
fi ( t 1) (5.4)
^
f in n
Mni fin (t ) , f Mni f fin (t )
where f is represented as objective function, t is the current iteration, n is

the number of feature vectors, and βn is the nth feature vector. The intensity
value and row and column position scores are computed for each pixel in
the fixed region (nose, right eye, left eye, and mouth). The neutral image as
well as the emotion image is used to calculate the feature quality. For the
purpose of training, the proposed hybrid network is provided the feature
values for all of the images.
5.5 Proposed HDC-GEN Classification

Classification is the process of predicting specific outcomes based on a set
of inputs. The method uses a training set, which consists of a collection of
qualities and their related outcomes, sometimes referred to as the target
or prediction attribute, to predict the outcome. The program tries to fig-
ure out if there are any correlations between the qualities that might help
predict the outcome. The HDC-GEN layer consists of three layers named
the input layer, output layer, and hidden layer. The HDC-GEN algorithm is
the combination of improved deep CNN and Golden Eagle Optimization.
A learning function in the hidden units is operated by the golden eagle
fitness function.
Initialization: Initialize the featured data as (aj, bj, cj). The kernel func-
tion is executed to the input function using Equation 5.5:
+∞
z=
(t ) =
m(t ) ∗ n(t ) ∑ m(τ )n(t − δ )
δ = −∞
(5.5)
where m is the input image feature and n is the convolution kernel perfor-
mance, τ is denoted as time delay, and t is symbolized as time, respectively.
The tangent hyperplane is evaluated for the kernel exploration by Equation
5.6:
a
h1c1 + h2c2 + ..haca= n ⇒ ∑h c = n
i =1
i i (5.6)
where h1, h2,……ha are the normal vector and c1, c2,……ca are the ith node
decision vector.
Max-Pooling Function: Once the kernel function is applied, then the
max-pooling operation is executed. The invariant features are extracted
in the kernel layer using the max-pooling function, turning the extracted
features into various images. The max-pooling function is applied in all
portions of the featured images using Equation 5.7:
x zj ,a = max
= p 1 (x z −1, a
( j −1)×o + p ) (5.7)
where o is denoted as the section range and p is considered as the width

window.
Optimized Fully Connected Training Layer: The fully connected layer
of HDC-GEN is used to train the non-linear combination of data. In
addition, the images are flattened into column vectors. For each train-
ing iteration, it is then fed into a feed-forward based deep net as well
as a backward propagation model. The training set is teaching the right
class size in this stage. Thus, the feed-forward propagation is evaluated
in Equation 5.8:
x zj = ∑h
a
α ( x zj1−1,a ) + baz −1
z −1
k, j (5.8)
The training may be characterized using a vector that starts at the cur-
rent position of the data and ends at the ideal weight points in the data
memory. The training vector for classification is computed using Equation
5.9:
Start
Initialize the different human emotion data
Apply median filter for preprocessing
Compute face detection function using

knowledge based approach
Apply face cropping function for image

resizing
Extract the features from the image using Initialized the features to
HBO method HDC-GEN for classification
Contribute the backward propagation in the

optimized fully connected training phase Apply max-pooling function
Compute the softmax operation for

classification
No
If criteria met? New emotion exploitation is
apply
Yes
Optimal human emotion is recognized
Stop
Figure 5.2 Flowchart for proposed human emotion recognition.

t az ≡ b zfa−1 − baz −1 (5.9)
where t az is the training exploitation vector, b z−1

fa is the best location of
data visited for the emotion, and baz−1 is the present position of the feature
point. Furthermore, the trained data is regularized by the Softmax opera-
tion for the further function of classification.
Classification: Based on the probabilities from the optimized training
stage, the emotions are classified further using Equation 5.10:
t
e x za
= =
f (x ) arg max m y(h | x ) = arg max m N
(5.10)
∑e
p =1
xt z p
where the weight vector is denoted as z, y is represented as the label vari-

able, the features of sample are considered as x, and the label of class is
denoted as m, respectively. As a result, the actions are classified using the
trained network. Once the best feeling has been identified, the criterion
ceases to work until the best result has been achieved. Finally, the sug-
gested approach’s performance is assessed using several metrics. Figure
5.2 shows the flowchart for the suggested approach of human emotion
recognition.
5.6 Result and Discussion

All of the applications were tested on a Windows platform with an Intel
i5 powered by an octa processor and 8 GB of RAM. Then, MATLAB
2018a was used to conduct the experiment. Initially, the expanded Cohn–
Kanade expression sample was used to assess the algorithm’s efficiency. As
the amount of images in the collection improves, so does the accuracy.
The dataset was firstly divided into two parts: training set with 80% of the
data as well as a testing set with 20% of the data. Several of the training
networks were uploaded and supplied the whole testing set with a single
image at a moment during the validation process. This was a brand-new
image that the model had not seen before. In the beginning, the image
input to the model was preprocessed. As a result, the model did not know
what the proper output was and had to forecast it accurately based on its
own training. It preceded to categorize the emotion depicted in the image
only based about what it had previously learned and the image’s features.
As a result, for each image, it generated a list of identified emotion proba-
bilities. The number of correct forecasts was calculated by comparing the
highest probability reaction for each image with the real feelings associated
with the visuals.
5.6.1 Performance Metrics Evaluation

The performance of the proposed method is validated with the subsequent
metrics such as accuracy, precision, F-measure, recall, error, and process-
ing time.
Precision: The precision metrics are one of the significant evaluation
measures for the performance analysis of proposed emotion classification.
This is evaluated by the ratio of particular predicted positive emotions to
the overall predicted positive emotion, which is expressed using Equation
5.11:
Tp
Precision = ∧ ∧
(5.11)
T p+ Fp
∧
where the true positive is denoted as T p and false positive is signified as

∧
F p.
Accuracy: The accuracy is another important parameter for the evalua-
tion measures in emotion classification. It characterizes the accurate per-
centage of classified emotion and when the accuracy reaches 100%, then it
is referred to as the classification of emotion at its finest and is estimated
using Equation 5.12:
∧ ∧
T p+Tn
Accuracy = ∧ ∧ ∧ ∧
(5.12)
T p +Tn + F p + Fn
Recall: It defines the particular amount of accurate positives and is clas-

sified particularly as optimistic, expressed using Equation 5.13:
Tp
Sensitivity = ∧ ∧
(5.13)
T p + Fn
F-Measure: The metrics of F-measure are estimated by the ratio of mean

weight of precision and recall using Equation 5.14:
∧
2T p
F-Measure = ∧ ∧ ∧ (5.14)
2T p + F p + F n
Error Rate: The proposed emotion classification error rate is evaluated

using Equation 5.15 as:
Error = 1 − Accuracy (5.15)
5.6.2 Comparative Analysis

The performance measures attained from the proposed approach are
compared with the traditional models GPSO-CNN [15], BLSTM-RNN
[13], and SSA-DCNN in terms of precision, accuracy, recall, error rate,
F-measure, and processing time. The precision value of the proposed
approach for anger, sadness, fear, surprise, and happiness is compared
with the conventional methods, GPSO-CNN [15], BLSTM-RNN [13] and
SSA-DCNN, and is illustrated in Figure 5.3. From this, the happy emotion
achieved the highest precision value by the proposed approach (98.9%)
when compared with the existing methods GPSO-CNN (98%), BLSTM-
RNN (74%), and SSA-DCNN (96.45%).
100
80
GPSO-CNN
Precision (%)
60 BLSTM-RNN
SSA-DCNN
Proposed
40
20
0
Angry Sadness Fear Surprise Happy
Figure 5.3 Comparison of precision over existing methods.

The accuracy value of the proposed approach for anger, sadness, fear,
surprise, and happiness is compared with the existing methods GPSO-
CNN [15], BLSTM-RNN [13], and SSA-DCNN and is illustrated in
Figure 5.4. From this, the happy emotion achieved the highest accuracy
value by the proposed approach (99.2%) when compared with the exist-
ing methods GPSO-CNN (69%), BLSTM-RNN (86%), and SSA-DCNN
(95.33%).
Moreover, the recall value of the proposed approach for anger, sadness,
fear, surprise, and happiness is compared with existing methods such as
GPSO-CNN [15], BLSTM-RNN [13], and SSA-DCNN and is illustrated
in Figure 5.5. From this, the happy emotion achieved the highest recall
value by the proposed method (97.6%) when compared with the exist-
ing methods GPSO-CNN (80%), BLSTM-RNN (88%), and SSA-DCNN
(96%).
The F-measure value for the five different emotions compared with the
existing methods such as GPSO-CNN [15], BLSTM-RNN [13], and SSA-
DCNN is illustrated in Figure 5.6. From this, the happy emotion achieved
the highest F-measure value by the proposed method (97.98%) when com-
pared with the existing methods GPSO-CNN (85%), BLSTM-RNN (88%),
and SSA-DCNN (95%).
Also, the estimation of processing time is the foremost measure in this
study because less processing time is more significant than high time. The
processing time for the five different emotions compared with the existing
100
80
Accuracy (%)
60
GPSO-CNN
40 BLSTM-RNN
SSA-DCNN
Proposed
20
0
Figure 5.4 Comparison of accuracy over existing models.

100
80
60
Recall (%)
GPSO-CNN
40 BLSTM-RNN
SSA-DCNN
Proposed
20
0
Figure 5.5 Comparison of recall with existing methods.
Proposed
Happy
SSA-DCNN
BLSTM-RNN
Surprise GPSO-CNN
Fear
Sadness
Angry
0 20 40 60 80 100
F-measure (%)
Figure 5.6 Comparison of F-measure over conventional models.
methods such as GPSO-CNN [15], BLSTM-RNN [13], and SSA-DCNN

is illustrated in Figure 5.7. From this, the happy emotion achieved less
processing time by the proposed method (14ms) compared with the exist-
ing methods GPSO-CNN (60ms), BLSTM-RNN (35ms), and SSA-DCNN
(50ms).
60
Processing time (ms)
40
GPSO-CNN
BLSTM-RNN
SSA-DCNN
20
Proposed
0
Figure 5.7 Comparison of processing time over existing models.
The error value obtained from the proposed classification model for
the five different emotions is compared with the existing methods such as
GPSO-CNN [15], BLSTM-RNN, [13] and SSA-DCNN and illustrated in
Figure 5.8. From this, the happy emotion achieved less error than the pro-
posed method (0.01%) compared with the existing methods GPSO-CNN
(1%), BLSTM-RNN (0.2%), and SSA-DCNN (0.18%).
1.8 GPSO-CNN
BLSTM-RNN
1.6 SSA-DCNN
Proposed
1.4
1.2
Error rate (%)
1.0
0.8
0.6
0.4
0.2
0.0
Figure 5.8 Comparison of error rate over conventional approaches.

5.7 Conclusion
Emotion recognition is a difficult problem that is rising in relevance due
to its numerous applications. Facial expressions can be useful in assessing
a human’s emotion or mental state. The proposed design solutions were
executed effectively and a computational simulation verified their validity.
The development of a hybrid model and a program for identifying emo-
tions by facial expression is the major result of this study. Each category
of emotions has its own set of recognition accuracy metrics. Images with
the emotion “happy” (99.2%) have the best recognition accuracy, while
images with the emotion “sadness” have the worst (97%). Also, the com-
parative analysis proves that the proposed method has achieved superior
performance in emotion recognition over the existing models in terms of
high accuracy, precision, F-measure, recall, and less processing time as well
as error rate. In the future, a new intelligent algorithm can develop with
proper validation of various human emotions.
References
1. Ghanem, Bilal, Paolo Rosso, and Francisco Rangel. “An emotional analysis of
false information in social media and news articles.” ACM Transactions on
Internet Technology (TOIT) 20.2 (2020): 1-18.
2. Dasgupta, Poorna Banerjee. “Detection and analysis of human emo-
tions through voice and speech pattern processing.” arXiv preprint
arXiv:1710.10198 (2017).
3. Chen, Caihua. “An analysis of Mandarin emotional tendency recognition
based on expression spatiotemporal feature recognition.” International
Journal of Biometrics 13.2-3 (2021): 211-228.
4. Sánchez-Gordón, Mary, and Ricardo Colomo-Palacios. “Taking the emo-
tional pulse of software engineering—A systematic literature review of
empirical studies.” Information and Software Technology 115 (2019): 23-43.
5. Hemanth, D. Jude, and J. Anitha. “Brain signal based human emo-
tion analysis by circular back propagation and Deep Kohonen Neural
Networks.” Computers & Electrical Engineering 68 (2018): 170-180.
6. Du, Guanglong, et al. “A convolution bidirectional long short-term mem-
ory neural network for driver emotion recognition.” IEEE Transactions on
Intelligent Transportation Systems (2020).
7. Ashwin, T. S., and Guddeti Ram Mohana Reddy. “Automatic detection of stu-
dents’ affective states in classroom environment using hybrid convolutional
neural networks.” Education and Information Technologies 25.2 (2020):
1387-1415.
8. Nonis, Francesca, et al. “Understanding Abstraction in Deep CNN: An

Application on Facial Emotion Recognition.” Progresses in Artificial
Intelligence and Neural Systems. Springer, Singapore, 2021. 281-290.
9. Sarmah, Dipti Kapoor. “A survey on the latest development of machine learn-
ing in genetic algorithm and particle swarm optimization.” Optimization in
Machine Learning and Applications. Springer, Singapore, 2020. 91-112.
10. Liu, Junjie. “Automatic Film Label Acquisition Method Based on
Improved Neural Networks Optimized by Mutation Ant Colony
Algorithm.” Computational Intelligence and Neuroscience 2021 (2021).
11. Mehendale, Ninad. “Facial emotion recognition using convolutional neural
networks (FERC).” SN Applied Sciences 2.3 (2020): 1-8.
12. Mungra, Dhara, et al. “PRATIT: a CNN-based emotion recognition system
using histogram equalization and data augmentation.” Multimedia Tools and
Applications 79.3 (2020): 2285-2307.
13. Li, Chao, et al. “Exploring temporal representations by leveraging atten-
tion-based bidirectional LSTM-RNNs for multi-modal emotion recogni-
tion.” Information Processing & Management 57.3 (2020): 102185.
14. Estrada, María Lucía Barrón, et al. “Opinion mining and emotion recognition
applied to learning environments.” Expert Systems with Applications 150
(2020): 113265.
15. Gao, Zhongke, et al. “A GPSO-optimized convolutional neural networks for
EEG-based emotion recognition.” Neurocomputing 380 (2020): 225-235.
6
An Application of Information
Technology in Adaptive Leadership of
Ministry of Ayush During Pandemic
of Covid 19: A Case Study
Vikram Singh1, Shikha Kapoor1 and Sandeep Kumar Gupta2*
1
AIBS, Amity University, Noida, India
2
AMET University, Chennai, India
Abstract
The pandemic of COVID-19, since its beginning, has been impacting the lives of
people the world over. However, in India, the impact would have been catastrophic
had Ayush’s intervention not been there. This paper examines the initiatives and
interventions by the leadership of the Ministry of Ayush (MoAyush) to manage
the pandemic situation through adaptive leadership. Effective communication,
use of IT platform, HR training, engagement with stakeholders, and adaptation
of R&D activities are the factors that helped the MoAyush to establish itself as the
leader in the fight against the pandemic of COVID-19.
Keywords: COVID-19, leadership, intervention, stakeholders, communication,

training, R&D
6.1 Introduction
The very first case of COVID-19 was detected in China on 12th December
2019. A comprehensive package of technical guidance was issued by the
World Health Organization (WHO) with advice to all countries about
detection, testing, and managing potential cases on 10th January 2020. The
Govt. of India immediately directed the states and UTs to take all nec-
essary action for health sector preparedness following WHO’s advisory.
*Corresponding author: skguptabhu@gmail.com

77
The screening of incoming passengers at the airports of Delhi, Mumbai,

and Kolkatta commenced on the 17th January 2020 and further extended to
airports at Chennai, Cochin, Bengaluru, and Hyderabad on 21st Jan 2020.
The first COVID-19 positive case in the country was detected on 30th Jan
2020 in Kerala.
During times of crisis, there is a generally widespread sense of volatil-
ity, uncertainty, complexity, and ambiguity which need very prompt and
high impact decisions in a limited information setting. Leaders have to
manage these elements and adapt to new situations along the way. The cri-
sis management is the set of activities targeted at reducing the negative
consequence of crises. Effective crisis management saves lives, safeguards
infrastructure, and restores public faith (Boin et al., 2013). Several charac-
teristics and practices that effective leaders embody during the time of cri-
ses are communication, decision making, humanism, innovation, realism,
and core values (Kaul et al., 2020). This paper examines the main charac-
teristics and practices the MoAyush adopted during the crises.
The Ministry of Ayush, while taking stock of the situation, issued an
advisory on ways to protect the general public from COVID-19 and stay
healthy (PIB release 29th Jan 2020). The advisory comprised of simple
home remedies, including the use of common Ayurvedic, Homeopathic,
and Unani medicines.
The Ayush Ministry kept a close watch on the pandemic spread and
responded with enhanced vigour and foresightedness. It took various
initiatives including an awareness generation campaign, Research and
Development (R&D), and pieces of training of Ayush manpower in clinical
practices to handle the future eventualities, etc. Various important initia-
tives have helped mitigate the impact that COVID-19 could have brought
to the nation.
6.2 Ministry of AYUSH

On the 9th of November 2014, the Department of Ayush was converted
into a full-fledged Ministry by the Government. It was done with a vision
for reviving the profound knowledge of ancient systems of medicine and
ensuring the optimal development and propagation of the Ayush sys-
tems of healthcare. The budget of the Department of Ayush was around
Rs 10 crore in 1992 and Rs. 1272 crores in 2014-2015, which became to
Rs. 2970.30 crores for the year 2021-2022. There has been a significant
increase in the work undertaken by the Ministry over the years. It is worth
mentioning that the budget of the Ayush Ministry is a small fraction of
Application of IT in Adaptive Leadership of MoAyush 79
the health budget but the impact among stakeholders has been large (PIB
release 7th Feb 2021).
The present government is very supportive of Ayush systems. Various
new and innovative programmes have been launched since 2014.
International Day of Yoga (IDY), the establishment of IT infrastructure in
the form of AyushGrid, the launch of National Ayush Mission, and certi-
fication of yoga professionals and accreditation of Yoga Institutions were
some of the significant contributions of the Ministry apart from regular
Central Sector Schemes. The Ayush services are available in all major gov-
ernment hospitals and Ayush treatment is part of the insurance policy now.
Out of 125,000 Health and Wellness Centres (HWCs) the government is
planning to establish, 10% would be dedicated to Ayush systems.
6.3 Leadership Principles and Practices by Ministry

of AYUSH During Covid-19
In the changing environment, the study of leadership during a crisis has
become more relevant because crises are more unpredictable, longer-
lasting, and costly than in the past (Pinsdorf, 2004). Leaders all across the
world are grappling with the difficulties of anticipating, responding to, and
trying to recover from the pandemic of COVID-19.
The pandemic of COVID-19 impacted every sector, but it put the health
sector to the test. The foundations of India’s healthcare have been shaken.
The public sector performed miserably compared to the private sector.
However, the Ayush sector came to the rescue during this crisis through
timely intervention and adaptive leadership. The following principles and
practices are worth noticing.
6.4 Effective Communication

Effective communication is very crucial during any crisis. Communication
done with urgency, transparency, and empathy is effective. It helps the
public to adjust to the tremendously changing conditions during any cri-
sis like the pandemic of COVID-19 has shown. The Ministry of Ayush,
without waiting for the widespread of Corona Virus, provided effective
communication in the form of press releases and advisories to Chief
Secretaries of the States/Union Territories and various stakeholders in
January 2020.
The Ministry launched a dashboard on its website which provided

COVID-19 related information in English as well as in Hindi. Immunity
boosting measures for subject care were placed at a very conspicuous place
and visible immediately on the landing page. The Ministry also opened
an outlet for an immunity booster kit at its Headquarters and also made
it available on the Amazon online platform. One of the aspects of effec-
tive communication during a crisis is the efficacy of communication. The
Ministry through its website, social media, and autonomous bodies reg-
ularly communicated with the stakeholders and public at large. Various
campaigns like “Ayush for Immunity” and “Yoga for Wellness” were
launched on a mass scale. The Ministry also invited suggestions and pro-
posals from various organizations under its Extra Mural Research Scheme
to fight COVID-19. Hundreds of proposals were received and considered.
Further, MoA also issued directives to all State/UT licensing authori-
ties and drug controllers of Ayush to increase the production of immunity
booster medicines. On one side, efforts were on for promoting immunity
efforts but some unscrupulous people were advertising and promoting
misleading information about Ayush intervention and were adverting
claims for treatment of Covid-19 and cheating people. The ministry took
this misleading information very seriously and directed all Regulatory
Authorities in the States/Union Territories, through its circular dated 1st
April 2020, to stop and prevent publicity and advertisement of Ayush-
related claims for COVID-19 treatment in print, TV, and electronic media
and to take necessary action against the persons/agencies involved in con-
travening the relevant legal provisions and the aforesaid National Disaster
Management guidelines (NMMA).
6.5 Sharing of Resources

COVID-19 drastically changed the way people work, shop, eat, travel,
and meet. The pandemic set up a paradigm in employment landscaping
that required re-scaling and re-training. The Ayush practitioners, though
equally competent, were only a fraction of those engaged in clinical man-
agement. With conventional health infrastructure coming under huge
stress, Ayush infrastructure provided the balance in supply and demand
for clinical management. The Ministry enabled the availability of infra-
structure in all Ayurveda, Siddha, Unani, and Homoeopathy institutes to
act like hospitals, labs, and ICUs along with medical and Para-medical
staff to compete with the COVID-19 pandemic. The States and the UTs
were suggested to use Ayush infrastructure, which comprises of 50000
beds attached to 750 Ayush Colleges and 86 clinical facilities and Research
Councils under the Ministry. All the hospitals, colleges, centres, and facil-
ities were shared.
6.6 Shared Decision Making

Shared decision making involves the bidirectional exchange of information
and values, better implementation of the decisions, and enhancement of
management strategies (Abrams et al., 2020). Before all the advisories and
directions, MoAyush not only consulted the stakeholders but also made
shared decisions. The India Institute of Ayurveda (AIIA) launched the
Ayuraksha programme for Delhi Police and their families (PIB, 30th April
2020). All the research councils and national institutes were consulted by
the Ministry on regular basis. Ayush-64, an Ayurveda medicine found effec-
tive in mild and moderate cases, was freely distributed for COVID-19 pos-
itive cases in the second wave. Further, as part of AzadikaAmrutMahotsav,
the immunity booster medicines would be distributed to 75 lakh people
across the country, with special attention on the geriatric population and
front line workers (PIB release 2nd Sept 2021).
6.7 Training of Manpower

The Ministry of Ayush and the Ministry of Health and Family Welfare
jointly organized training of around 33000 Ayush Master Trainers for fur-
ther training of the Ayush workforce, considering the future requirement
to manage the pandemic in April 2020. Over 66000 Ayush personnel were
trained at the Integrated Government Online Training (iGOT) portal. The
effectiveness of training on iGOT can be seen in that over 37000 Ayush staff
were successfully deployed for COVID-19 related health services. In order
to attain a high degree of uniformity in sickness care, the ministry has set
rules for registered practitioners of distinct Ayush systems (Guidelines for
Ayush Practitioners for COVID 19).
6.8 Use of IT Platform

The complete lockdown was announced in the country on 24th March
2020, commencing at midnight. All services including healthcare services
in OPD were at a standstill. There was an urgent need for guidelines on
Telemedicine for Ayush systems to the community of service. The ministry

issued the guidelines on 7th April 2020 on its website.
On May 7, 2020, the Ministry launched the “Ayush Sanjeevani” mobile
app to collect data on experience, usage of Ayush advocacies within the
population, and its impact on COVID-19 prevention. This app would
gather information on the measures applied by the people for enhancing
immunity and keeping themselves healthy during COVID-19 situations
(PIB release dated 7th May 2020). Data from over 723 thousand respondents
for three months, May-July 2020, revealed that 85.2% used Ayush methods
for maintenance of body immunity during the COVID-19 pandemic and
around 89.8% of them benefitted from AYUSH measures. (Srikanth et al.,
2021).
Further, the Ayush Clinical Case Repository (ACCR) portal (https://
accr.ayush.gov.in) was established and developed by the Ministry as a plat-
form to support both Ayush practitioners and the public to portray the
strength of the Ayush System. In this portal, Ayush practitioners from any-
where in the world can enrol themselves and share relevant information
about their successful treatment for the benefit of all. It was launched on
27th May 2021 (PIB release dated 26th May 2021). As of now, it has 28 cases
including 12 on COVID-19.
The Ministry of Ayush already initiated the Ayush Grid Project in the
year 2018 for creating a comprehensive IT backbone for the entire Ayush
Sector (Yadava & Ps, n.d.). Through the Ayush Grid, the entire Ayush
Sector would be digitalized and it would lead to sector transformation in
the field of healthcare delivery at all levels and would help research, edu-
cation, drug development, and regulation of various health programmes
of the Government. The Ayush Grid would be beneficial for all the stake-
holders of Ayush at the National and Global levels. Under the Ayush
Grid, the project of the Ayush Health Management Information System
(AHMIS) had been developed into a quality based information system and
over 100 clinical establishments of the Ministry are using it successfully.
The Ministry also offered the AHMIS to all the Ayush units outside the
Government umbrella so that the entire Ayush Sector benefits. The compo-
nents of Ayush Grid initiatives would gather and connect all the folks and
stakeholders of the Ayush Sector. The Ayush Grid is expected to emerge
as a comprehensive IT backbone serving 0.85 million Ayush physicians
and over 500 million citizens within 3 years. The Ayush Grid has been
integrated with the National Digital Health Mission (NDHM) and will be
useful to the public for varied options of health needs. The integration of
the Ayush Grid into NDHM will speed up the mainstreaming of Ayush
systems of healthcare (PIB release dated 2nd Oct 2020).
The Ministry also launched a dedicated yoga portal (https://yoga.ayush.

gov.in/) to meet the demand of Yoga in the country as well as abroad.
The portal contains information on Common Yoga Protocol (CYP), the
audiovisual source material for training and practice on CYP, and informa-
tion related to IDY. Further, a Yoga Break (Y-Break) mobile app was also
launched for practising 5 minutes of yoga at the workplace. The protocol
has a set of stretches, pranayama, and meditation developed by experts.
Y-break, when practised twice a day, would reduce stress and refresh and
refocus the workforce.
A pilot project of Telemedicine was launched in the Siddha System of
medicine in 100 villages in Tamil Nadu in Nov 2019 by the Ministry with
the help of Common Service Centres. Detailed guidelines of telemetry
of Ayurveda, Unani, and Siddha practitioners were issued by the Central
Council for Indian Medicines on 7th April 2020 and these guidelines are
also made available on the website of the Ministry in line with India’s digital
health policy. The Ministry of Ayush also developed a dashboard (https://
dashboard.ayush.gov.in/) for various information in the Ayush Sector like
Ayush health infrastructure, health, drug industry, institute, research, bud-
get, etc. Another project launched by the Ministry of Ayush is “Ayush Next”
to support Ayush education. It is a digital platform intended to enhance
the realm of information exchange by providing career recommendations,
an interactive forum, quizzes, and more. The group of professionals and
domain experts is well-versed in resolving concerns about careers and
employment in the Ayush sector. Information on various degrees and cer-
tifications is also available. It is a place for young people of the Ayush sec-
tor to share their thoughts and knowledge, seek guidance and counselling,
discuss, and continue to learn.
6.9 Finding Opportunities for R&D During the Crisis

Ayush Systems have been criticized for not being practically scientific
and lacking evidence. The pandemic brought a different type of oppor-
tunity before the Ayush Sector to collect fresh evidence and re-establish
itself with modern medical science. An Interdisciplinary Ayush R&D Task
Force was constituted with Prof. Bhushan Patwardhan as chairperson and
other members from the Council of Scientific and Industrial Research
(CSIR), the Indian Council of Medical Research (ICMR), Department of
Biotechnology (DBT), All India Institute of Medical Sciences (AIIMS),
and prominent Ayush Institutions in the first week of April 2020. It
designed and formulated clinical research protocols for prophylactic
studies and add-on interventions in COVID-19 positive cases. Further,

The Ministry of Ayush solicited research ideas from diverse stakeholders
in order to develop scientific proof. Over 125 clinical research projects
and basic experimental investigations were initiated in over 150 centres
across the country based on the recommendations of the Interdisciplinary
task group. To name a few, these studies were conducted by King George
Medical University Lucknow, All India Medical Institute New Delhi,
Institute of Medical Sciences BHU, Govt. Medical College Nagpur,
KEM Hospital Pune, National Institute of Pharmaceutical Education
and Research Kolkatta, DMIMS Wardha Indian Institute of Integrative
Medicine Jammu, and Institute of Genomics and Integrative Biology
(IGIB) New Delhi. In addition, the Central Council for Study in Yoga
and Naturopathy (CCRYN) has initiated a research initiative to validate
the efficacy of Yoga in supporting early recovery of COVID– 9 patients in
collaboration with Rajiv Gandhi Super Speciality Hospital, Delhi, AIIMS
Delhi, AIIMS Rishikesh, and RML Hospital, Delhi.
The All India Institute of Ayurveda (AIIA) discovered that Ayurvedic
therapies such as Ayush kwatha and Fifatrol pills might be helpful in mild
to severe COVID-19 infection in a “quite short period” with “total regres-
sion of symptoms.” As per the case report published in the AIIA journal
‘Ayurveda Case Report’ in October 2021, the use of four Ayurvedic inter-
ventions, Sanshamanivati, Ayush kwatha, Laxmivilasa rasa, and Fifatrol
tablets, not only improved the condition of the COVID-19 infected patient
but also gave a negative report on the fast antigen test within six days of
medication.
6.10 Collaborating with Stakeholders

for International Day of Yoga (IDY)
The United Nations General Assembly (UNGA) unanimously recognised
21st June as the International Day of Yoga (IDY) on the appeal of PM
Narendra Modi. The day has been celebrated around the world in ever-
increasing numbers. The Ministry of Ayush is the nodal ministry respon-
sible for managing and promoting the practice of Yoga and coordination
activities for the observation of IDY every year. The IDY observance for
the year 2021 could not be more timely, as the world and India at large reel
from a deadly wave of the pandemic as new variants emerged. Hence, in
that spirit, the theme of this year was kept as “Yoga for Wellness” to remind
the world about Yoga and the holistic approach to health and wellbeing it
takes.
Despite IDY observances, though overcast by restrictions due to the
pandemic for a second time, the Ministry learned from previous years’
experience. The Ministry followed a three-pronged strategy which
included taking a digital-first approach, activating Government of India
stakeholders and their networks which have a wide-ranging reach, and
collaborating with the private sector in a greater way to pave the road for
the adoption of Yoga at a larger scale and development of the sector in
the years to come.
In the digital-first approach, the Ministry leveraged its existing IT
platforms and social media including Facebook, Twitter, Instagram,
YouTube, the Ministry’s website, and the Yoga Portal; it also leveraged
new platforms such as Koo, Josh, and Bolo India to reach wider viewers
which consume diverse kinds of content and prefer local and regional
languages. The Ministry released additional digital assets apart from
improvements in the existing ones, namely the Namaste Yoga appli-
cation. Further, the Government’s MyGov platform was used for the
first time to launch citizen-centred awareness building and engaging
activities. People were encouraged to practice yoga based on Common
Yoga Protocol (CYP) virtually across the country. Various IT tools and
online resources were created during the pandemic. Yoga has been very
useful for the well-being of people during the pandemic. The Prime
Minister of India himself leads the nation for IDY celebration. He
launched the “mYoga” application, developed in house by the Morarji
Desai National Institute of Yoga in collaboration with the World Health
Organisation (WHO) on 7th IDY on 21st June 2021. The app can be used
by all to promote one world, one health. (PIB release 21st June 2021).
The mYoga application focuses on WHO’s theme of ‘Be He@lthy, Be
Mobile (BHBM)’, which is in line with the United Nations’ Sustainable
Development Goal of achieving Universal Health Coverage by 2030.
The app addresses diseases and risk factors with an opportunity to nav-
igate upon the m-Health programme for Yoga. It offers a collection of
learning modules and practice sessions in both audio and video formats,
which enable users to practice Yoga anytime from the comfort of their
homes. The mYoga app is available in the Google Play Store and Apple
App Store. As per the AppBrain ranking report, (https://www.appbrain.
com/), the WHO mYoga app is placed at a Google Play store rating of
4.75. It has over 50+ thousand downloads.
The yoga portal of the ministry had a viewership of over 1.05 million.
The Yoga Certification Board (YCB) for the Yoga Appreciation Programme
and Yoga Volunteer Training (YVT) Course, reported an outreach of
about 1.214 million. The Prime Minister addressed the nation on IDY
through various DD channels, emphasising the role of maintaining health
and well-being in the lives of people, especially at a time when the world
was reeling from the COVID pandemic.
All the Central Ministries, State Governments, public sector under-
takings, autonomous bodies, various yoga organizations, Universities,
National Cadet Corps (NCC), National Social Service (NSS), Nehru
Yuva Kendra (NYKs), schools, hospitals, railways, postal services, and
GramPradhans were brought on board. To reach the nooks and corners
of the country, MoAyush also used services of various Common Service
Centres (CSC) and Community Radio Stations to promote IDY activities
and motivate people to practice yoga. The Ministry also had a partnership
with Nickelodeon and HealthifyMe to reach different segments of people.
Indian Missions Abroad were also equally active in the promotion and
observation of IDY. The coordinated efforts of MoAyush have made Yoga
a mass movement and around 156.86 million people participated during
IDY 2021 (Table 6.1).
Table 6.1 Social media outreach during International Day of

Yoga 2021.
Social media platform Outreach (in thousands)
YouTube 2,100
Facebook 11,551
Twitter 8,910
Instagram 4,100
KOO 6,000
JOSH 3,57,300
Bolo India 700
Total Social Media 3,90,661
Summary of Platform-wise Outreach of IDY 2021 (Source MoAyush).
6.11 Providing Hope When Nothing Seemed

to be Working
Ayush, through its communications, continued to inform the citizens to use
Ayush products, particularly home remedies for improving immunity in the
fight against COVID-19. This hope has helped millions, which is evident from
the demand for Ayush immunity booster products, including Chavanprash.
The social media of the Ayush Ministry kept engaging people on the use
of Ayush products. Prime Minister spoke on Ayush Kadha in mannkibaat.
The Secretary of the Ministry, Rajesh Kotecha, who himself is a prominent
Ayurveda expert and a Padma Shri awardee, on almost all forums emphasized
that Ayush systems were effective tools in the management of COVID-19.
Yoga sessions were arranged in Covid Care Centres and were well received.
6.12 Leveraging Old Knowledge

Yoga and Naturopathy have been known for their long term positive impact
on the body, mind, and spirit. In the lockdown period, commencing in
March 2020, more than physical fitness, the mental and emotional health
of people was put to the test. Yoga came as a saviour. The Ministry of Ayush
through its autonomous bodies, the Morarji Desai National Institute of
Yoga, and the National Institute of Naturopathy (NIN), conducted regular
yoga sessions for health and wellbeing. Further considering the impact on
mental well-being, the Ministry of Ayush in collaboration with SVYASA
and NIMHANS also released a protocol for psychosocial rehabilitation of
COVID-19 infected patients.
In October 2020, the Ayush Ministry issued suggestions to incorpo-
rate “Ayurveda and Yoga therapies” into India’s national clinical manage-
ment plan for COVID-19 infection. It suggested treating COVID-19 with
warm water gargles, medicinal ghee in nostrils, steam inhalation, drink-
ing “golden milk” (turmeric mixed with hot milk), and kashayam/kadha/
kwath (hot infusion with Ayurvedic herbs) in addition to a healthy diet,
sleep, and exercise. The clinical protocol advised patients to take Ayurvedic
formulations such as Guduchi and Guduchi Ghana Vati, with Asvagandha
and Pipalli, even if they were suffering from hypoxia (a lack of oxygen in
the body) and dyspnea caused by COVID-19 infection.
6.13 Conclusion
The crisis of COVID-19 has impacted the whole country, but for the Ayush
sector, it brought tremendous opportunities. The ministry of Ayush pro-
vided timely intervention and has helped reposition Ayush therapy not
only at the national level but worldwide. It was made possible with IT
tools and the use of the digital platform. Effective and regular communi-
cation during the continuity of the crises, sharing of resources with other
stakeholders, use of IT platform regressively, training and preparedness of
existing manpower, and leveraging old wisdom were some of the practices
which helped MoAyush to establish itself as a true leader at the time of
crisis.
References
1. Abrams, E. M., Shaker, M., Oppenheimer, J., Davis, R. S., Bukstein, D. A.,
&Greenhawt, M. (2020). The Challenges and Opportunities for Shared
Decision Making Highlighted by COVID-19. In Journal of Allergy and
Clinical Immunology: In Practice. American Academy of Allergy, Asthma
and Immunology. https://doi.org/10.1016/j.jaip.2020.07.003.
2. Boin, A., Kuipers, S., & Overdijk, W. (2013). Leadership in times of crisis:
A framework for assessment. International Review of Public Administration,
18(1), 79–91. https://doi.org/10.1080/12294659.2013.10805241
3. Kaul, V., Shah, V. H., & El-Serag, H. (2020). Leadership During Crisis:
Lessons and Applications from the COVID-19 Pandemic. Gastroenterology.
https://doi.org/10.1053/j.gastro.2020.04.076
4. Kaul V, Shah VH, EI-Serag H (2020). Leadership During Crisis: Lessons and
Applications from the COVID-19 Pandemic, Gastroenterology.
5. Lee A. Wuhan novel coronavirus (COVID-19): why global control is chal-
lenging? Vol. 179, Public Health. Elsevier B.V.; 2020. p. A1-2
6. Ministry of AYUSH, Govt. of India 2020. D.O. No. S.16030/18/2019-NAM
dated 6th March 2020, Ministry of AYUSH, Govt. of India 2020. Ayurveda’s
immunity-boosting measures for self-care during COVID 19 crisis. https://
www.ayush.gov.in/docs/123.pdf
7. Ministry of AYUSH, Govt. of India 2020.Guidelines for AYUSH Practitioners
for COVID 19. https://www.ayush.gov.in/ayush-guidelines.html
8. Pinsdorf, M. K. (2004). All Crises are Global: Managing to Escape
Chaos. United States: Fordham University Press.
9. Srikanth N, Rana R, Singhal R, Jameela S, Singh R, Khanduri S, Tripathi
A, Goel S, Chhatre L, Chandra A, Rao B, Dhiman K (2021). Mobile App–
Reported Use of Traditional Medicine for Maintenance of Health in India
During the COVID-19 Pandemic: Cross-sectional Questionnaire Study,

JMIRx Med 2021;2(2):e25703, URL: https://xmed.jmir.org/2021/2/e25703,
DOI: 10.2196/25703
10. Yadava, R., & Ps, A. (n.d.). Development and contributions of AYUSH sector -
A Review.
11. https://pib.gov.in/PressReleasePage.aspx?PRID=1621787 dated 7th May 2020
12. https://pib.gov.in/Pressreleaseshare.aspx?PRID=1695897 dated 7th Feb 2021
13. https://pib.gov.in/PressReleasePage.aspx?PRID=1619678 dated 30th April 2020
14. https://pib.gov.in/PressReleasePage.aspx?PRID=1721962 dated 26th May 2021
15. https://pib.gov.in/PressReleasePage.aspx?PRID=1729128dated 21st June 2021
16. https://pib.gov.in/PressReleaseIframePage.aspx?PRID=1751452 dated 2nd Sept
2021
17. https://www.appbrain.com/app/who-myoga-app/org.who.APPMYOGA
18. https://pib.gov.in/PressReleasePage.aspx?PRID=1600895, dated 29.1.2020
7
Encoder-Decoder Models for Protein
Secondary Structure Prediction
Ashish Kumar Sharma* and Rajeev Srivastava
Department of Computer Science and Engineering, Indian Institute of Technology

(BHU), Varanasi, Uttar Pradesh, India
Abstract
Proteins are arranged in a linear sequence due to peptide bonds. In proteins, a
peptide bond combines the amino group of one protein with the carboxyl group
of another protein. Protein secondary structure formation results from their bio-
physical and biochemical properties, like natural languages which depend on their
grammatical rule. So, the proposed model predicts a secondary structure from
protein primary sequences using the encoder-decoder based machine translation
method. The proposed model uses an encoder-decoder model based on long short-
term memory network. The proposed work uses training and testing performed on
available public datasets, namely CullPDB and data1199. The proposed model has
better Q3 accuracy of 84.87% and 87.39% for CullPDB and data1199, respectively.
Further, the proposed work was evaluated by comparing their performance with
other methods which predict secondary structure only from a single sequence. The
Encoder-Decoder Model for predicting secondary structure from a single primary
sequence is performing better than other single sequence-based methods.
Keywords: Protein structure prediction, amino acids sequence, proteomics,
one hot encoding, encoder-decoder, long short-term memory
7.1 Introduction
Protein is important biomolecules which help in developing the cells of
all living organisms. Proteins mainly perform the role of transporters,
catalysts, receptors, and hormones within all living organisms [1]. The
*Corresponding author: ashishkrsha.rs.cse17@itbhu.ac.in
91
primary sequences of proteins are a linear sequence of twenty naturally

occurring amino acids. The amino acids are represented by one or three
letter symbols called codons. Three levels of structure mainly describe the
Protein structure [2]. The primary structure is linearly connected amino
acids with peptide bonds. The protein secondary structure is defined by
the local segments which are categorized into eight classes. These eight
categories of secondary structure are the alpha-helix (H), 3-10 helix (G),
pi helix (I), beta-sheet (E), beta bridge (B), turn (T), bend (S), and loop
(L) [3]. Further, the eight classes of secondary structure are simplified
into three classes: helix, strand, and loop/coil. The tertiary structure is the
three-dimensional structure. To predict the secondary structure of a pro-
tein, the primary sequence is a sequence-labeling task. The protein second-
ary structure helps in estimating its functions, which is essential for drug
designing and protein engineering.
In natural language processing, there are several character-based meth-
ods for classification and prediction. These character-based methods are
mainly classified on the basis of their character-level information into
three categories. These three character-based models are bag-of-n-gram,
tokenization-based, and end-to-end [4]. The neural machine translation
has become popular with better performance of character-level models.
The encoder-decoder models, which use character-based deep neural net-
works, are applied to several problems such as machine translation, ques-
tion answering, and speech recognition [5].
The protein sequences are formed with amino acids which are defined
by biophysical and biochemical principles. The biophysical and biochemi-
cal principles of proteins are similar to the grammar of natural languages.
This similarity motivates protein sequences that are the output of a spe-
cific biological language and develops a method based on natural language
processing such as the encoder-decoder method to discover functions
encoded within protein sequences.
The contributions made by this work are: (1) in the proposed model,
the primary sequence map to the secondary structure sequence as lan-
guage translation; (2) one LSTM layer encodes the primary sequences
and returns its internal state to input in the decoder and another LSTM
layer works as a decoder for secondary structure prediction; and (3) the
proposed encoder–decoder model for secondary structure prediction
was evaluated on two datasets, cullpdb and data1199. The experiments
performed show that the proposed encoder-decoder model better cap-
tures the features from amino acid sequences to predict their secondary
structure.
Encoder-Decoder for Protein Secondary Structure Prediction 93
7.2 Literature Review

In the last few decades, various methods have been proposed which are map-
ping a protein’s primary sequence to their secondary structure sequence.
Protein secondary structure prediction methods combine sequence homol-
ogy searches [6–8] with basic features such as physicochemical properties of
the primary sequence [9] and backbone torsion angles [10]. Further, these
combined features are fed into neural networks [11, 12] or deep neural net-
works [13, 14] that produce secondary structures of protein. Recently, meth-
ods used for protein secondary structure prediction are PSIPRED [15] and
JPRED [16], which have an average Q3 accuracy of approximately 80–85%
for benchmark datasets. PSIPRED was the first secondary structure predic-
tion tool that uses the PSI-BLAST search to improve prediction accuracy.
PSIPRED utilize the protein database UNIREF90 to obtain sequence pro-
files and pass to the neural network with two-layers. JPRED integrates the
PSI-BLAST sequence profile and HMMer [17] sequence profiles to the neu-
ral network. The currently available tools which have Q3 accuracy around
85–90% include SSpro [18], DeepCNF [19], PORTER [11], and PSRSM
[20]. SSpro uses template data by utilizing sequence homology sequences.
If homology sequences are not available, then the neural network predicts
their secondary structure class. Two methods, DeepCNF and PORTER,
combine the protein sequence profiles with deep conditional neural fields
and convolutional neural networks. The protein secondary structure pre-
diction methods only from a single sequence are PSIPRED-Single and
SPIDER3-Single [21–24]. These methods utilize the information extracted
from a single sequence to find the protein secondary structure.
7.3 Experimental Work

7.3.1 Data Set
To train the model, protein data was obtained by downloading from cull-
pdb [25], which has a resolution value below 2.5 Å, R-factor below 1.0,
and 30% non-redundant sequences in February 2017. The sequences with
incomplete information, having a length value below 30 residues, and with
find similarity more than 30% are removed by BlastClust [26]. We used
publicly available dataset data1199 as a testing set. The test set is 1199
sequences with no redundancy [27].
7.3.2 Proposed Methodology

The proposed protein secondary structure prediction model is a long short-
term memory network-based encoder-decoder architecture. The input is
the primary sequence and the output is the protein’s secondary structure.
Naturally, both the input protein primary sequences and output secondary
structure sequences are of equal lengths since each amino acid character
has a corresponding secondary structure character. In the proposed model,
the conditional probability of a secondary structure sequence (y1, . . . , ym)
estimated is given as an input amino acid sequence (x1, . . . , xn).
p ( y1 , y2 ........ yn | x1 , x2 ,..........xn ) (7.1)
The proposed encoder-decoder model first encodes the input protein

primary sequence and one amino acid character in the primary sequence
is represented using a latent vector. Then, it decodes to a protein secondary
structure sequence.
7.3.3 Data Preprocessing

The protein sequences are split into twenty amino acid characters. Each
amino acid is numbered with an integer value in the range of 1-20. The
amino acid characters are represented as one-hot vector. The protein
sequences are of varying length, but the deep learning model accepts the
fixed length. If any sequence exceeds in size, then the remaining character
is discarded. We padded with zeros in the shorter sequence.
7.3.4 Long Short Term Memory

Long short term memory (LSTM) [28] uses the capability of input gate
(it), forget gate (ft), and output gate (ot) to control the flow of information
to operate selective read, selective forget, and selective write. To utilize the
information efficiently and discard the unnecessary information, the three
gates in LSTM use the current input, previous state, and output selectively.
The activation function depends on the gates used in the LSTM. The logis-
tic and sigmoid function are used as an activation function. The flow of
information depends on memory blocks used in the hidden layer. The gov-
erning equations for LSTM are as follows:
=at σ ( w xa .xt + uha .Outt −1 + ba ) (7.2)

=it σ ( w xi .xt + uhi .Outt −1 + bi ) (7.3)
=ft σ ( whf .xt + uhf .Outt −1 + b f ) (7.4)
=ot σ ( who .xt + uho .Outt −1 + bo ) (7.5)
state = atΘit + ftΘstatet−1 (7.6)
=
out statet ΘOt (7.7)
where Wxa, Wxi, Whf, and Who are weight matrices for input vectors xt, Uha,
Uhi, Uhf, and Uho, which are weight matrices for previous state Outt-1, ba,
bi, bf, and bo, which are the bias terms for each memory gate at, it, ft, ot. ʘ
signifies the Hadamard product (element-wise product) operation and σ is
the sigmoid function.
The encoder-decoder model for the primary protein sequence to the
secondary structure is depicted in the Figure 7.1. In the model, the encoder
and decoder both use LSTM having 256 units. The state vector represen-
tation obtained from the encoder layer works as the input to the decoder
y1 y2 y3 y4
Predicted
Secondary
Structure
Secondary Secondary Secondary Secondary Softmax

structure structure structure structure Function
Probability Probability Probability Probability
STATE VECTOR
Decorder
Encorder
h1 h2 h3 hfinal hinitial
One hot Encoding
0 0 0 0 0 1 0 0
0 0 1 0 0 0 0 1
Encoding
One hot
0 1 0 0 0 0 0 0
0 0 0 1 1 0 0 0
1 0 0 0 0 0 1 0
Supervised
x1 x2 x3 x4 Secondary y1 y2 y3 y4
Primary Sequences Structure
Figure 7.1 Encoder–decoder model for protein primary sequence to secondary structure.
LSTM. The encoder LSTM encoded the primary sequences, while the
second decoder LSTM predicts the secondary structure sequences. The
encoder LSTM layer processes the individual amino acid characters in
a sequence and finally, produce a state vector. The decoder receives the
state vectors from the encoder. After processing the complete input pri-
mary sequence, the decoder layer uses the secondary structure character to
decode the state vectors into a secondary structure sequence. The decoder
layer maximizes the value of log-likelihood of predicted secondary struc-
ture sequence on the basis of encoder layer state vectors and past protein
secondary structure characters.
The encoder layer processes the protein primary sequences by utiliz-
ing individual amino acid characters and finally, produce the state vectors,
while the decoder layer uses the secondary structure sequence and the
encoder layer state vectors.
7.4 Results and Discussion

The performance metrics from Q3 accuracy are used for evaluating sec-
ondary structure prediction. Python version 3.6.7 is used for implement-
ing the proposed sequence-to-sequence model. To implement the model at
the front end, keras is used while Tensorflow is used as the back end. Keras
is an open-source application programming interface for deep learning.
Tensorflow has an excellent computational ability.
The statistical method used to sample the dataset was 70 to 30 for the
training set and testing set to develop a generalized model. The encoder-
decoder model is based on a long short-term memory network, which has
several hidden layers and each layer has a number of processing cells, so
a large amount of data is used for processing. The RmsProp optimization
has a small minibatch of 64 to more quickly process the model. The weights
and biases are updated using categorical cross-entropy, which is the nega-
tive log likelihood loss.
The proposed encoder-decoder model performance compared with
other methods such as SPIDER3-Single [22] and PSIpred-Single [21], which
only use a single sequence for the cullpdb dataset, are listed in Table 7.1. We
find that the proposed model performance with single LSTM is higher than
other methods, i.e., SPIDER3-Single [22] and PSIpred-Single [21].
SPIDER3-Single combines the one-hot feature vector with Bidirectional
LSTM for protein secondary structure prediction and PSIpred-Single con-
siders statics of significant amino acid by calculating their correlation at
each segment.
Table 7.1 Comparison of performance of various single-sequence

based prediction on Cullpdb.
Methods Q3 (%)
Seq2Seq (LSTM) 84.87
SPIDER3-Single 73.24
PSIpred-Single 70.21
In order to show the effectiveness of the sequence-to-sequence model,

the proposed model performance is compared with single sequence based
methods SPIDER3 [22], JPred4 [29], and RaptorX [30] to predict protein
secondary structure for dataset data1199, are listed in Table 7.2.
Table 7.2 Comparison of various single-sequence

based prediction on data1199.
Methods Q3 (%)
Seq2Seq Model 87.39
SPIDER3 83.3
JPred4 79.3
RaptorX 81.5
Spider3 methods used a bidirectional long short-term memory network.

JPred4 used the JNet procedure and RaptorX used deep convolutional
neural fields for predicting the protein secondary structure. To estimate
the performance of our model, the protein sequences in the testing sets
and training sets have low similarity. The Q3 accuracy of our sequence-to-
sequence model is 87.39%.
7.5 Conclusion
In the proposed work, primary protein sequences are translated to their sec-
ondary structure using the sequence-to-sequence model. Protein primary
sequences are represented as one-hot encoding and directly feed to the
LSTM based sequence-to-sequence model. The proposed model has com-
paratively better performance on two publicly available datasets, CullPDB
and data1199. Despite its simplicity, the LSTM based sequence-to-sequence
model easily captures the complex relationship between amino acids and
their secondary structure.
References
1. Z. Wang, F. Zhao, J. Peng, J. Xu, Protein 8-class secondary structure predic-
tion using conditional neural fields., Proteomics. 11 (2011) 3786–92. https://
doi.org/10.1002/pmic.201100196.
2. Q. Jiang, X. Jin, S.-J. Lee, S. Yao, Protein secondary structure prediction:
A survey of the state of the art., J. Mol. Graph. Model. 76 (2017) 379–402.
https://doi.org/10.1016/j.jmgm.2017.07.015.
3. Y. Wang, H. Mao, Z. Yi, Protein secondary structure prediction by using deep
learning method, Knowledge-Based Syst. (2017). https://doi.org/10.1016/j.
knosys.2016.11.015.
4. H. Schütze, S. Schütze, H. Adel, E. Asgari, arXiv:1610.00479v3 [cs.CL] 1 May
2017, 2017.
5. E. Vylomova, T. Cohn, X. He, G. Haffari, Word Representation Models
for Morphologically Rich Languages in Neural Machine Translation,
in: Proc. First Work. Subword Character Lev. Model. NLP, Association
for Computational Linguistics (ACL), 2018: pp. 103–108. https://doi.
org/10.18653/v1/w17-4115.
6. A. Drozdetskiy, C. Cole, J. Procter, G.J. Barton, JPred4: A protein second-
ary structure prediction server, Nucleic Acids Res. 43 (2015) W389–W394.
https://doi.org/10.1093/nar/gkv332.
7. D. Ganguly, D. Roy, M. Mitra, G.J.F. Jones, Word Embedding based
Generalized Language Model for Information Retrieval, in: 2015. https://doi.
org/10.1145/2766462.2767780.
8. M. Ashburner, C.A. Ball, J.A. Blake, D. Botstein, H. Butler, J.M. Cherry, A.P.
Davis, K. Dolinski, S.S. Dwight, J.T. Eppig, M.A. Harris, D.P. Hill, L. Issel-
Tarver, A. Kasarskis, S. Lewis, J.C. Matese, J.E. Richardson, M. Ringwald, G.M.
Rubin, G. Sherlock, Gene ontology: Tool for the unification of biology, Nat.
Genet. (2000). https://doi.org/10.1038/75556.
9. J. Chen, G. Liu, V. Pantalone, Q. Zhong, Physicochemical properties of pro-
teins extracted from four new Tennessee soybean lines, J. Agric. Food Res. 2
(2020) 100022. https://doi.org/10.1016/j.jafr.2020.100022.
10. M.N. Faraggi, A. Arnau, V.M. Silkin, Role of band structure and local-field
effects in the low-energy collective electronic excitation spectra of 2H-NbSe
2, Phys. Rev. B - Condens. Matter Mater. Phys. 86 (2012) 035115. https://doi.
org/10.1103/PhysRevB.86.035115.
11. C. Mirabello, G. Pollastri, Porter, PaleAle 4.0: High-accuracy prediction of
protein secondary structure and relative solvent accessibility, Bioinformatics.
29 (2013) 2056–2058. https://doi.org/10.1093/bioinformatics/btt344.
12. A. Yaseen, Y. Li, Context-based features enhance protein secondary structure

prediction accuracy, J. Chem. Inf. Model. 54 (2014) 992–1002. https://doi.
org/10.1021/ci400647u.
13. R. Heffernan, K. Paliwal, J. Lyons, A. Dehzangi, A. Sharma, J. Wang, A. Sattar,
Y. Yang, Y. Zhou, Improving prediction of secondary structure, local back-
bone angles, and solvent accessible surface area of proteins by iterative deep
learning, Sci. Rep. 5 (2015) 1–11. https://doi.org/10.1038/srep11476.
14. S. Wang, J. Peng, J. Ma, J. Xu, Protein Secondary Structure Prediction Using
Deep Convolutional Neural Fields, Sci. Rep. 6 (2016). https://doi.org/10.1038/
srep18962.
15. L.J. McGuffin, K. Bryson, D.T. Jones, The PSIPRED protein structure pre-
diction server, Bioinformatics. 16 (2000) 404–405. https://doi.org/10.1093/
bioinformatics/16.4.404.
16. J.A. Cuff, M.E. Clamp, A.S. Siddiqui, M. Finlay, G.J. Barton, JPred: A consen-
sus secondary structure prediction server, Bioinformatics. 14 (1998) 892–
893. https://doi.org/10.1093/bioinformatics/14.10.892.
17. R.D. Finn, J. Clements, S.R. Eddy, HMMER web server: Interactive sequence
similarity searching, Nucleic Acids Res. 39 (2011) W29. https://doi.
org/10.1093/nar/gkr367.
18. C.N. Magnan, P. Baldi, SSpro/ACCpro 5: Almost perfect prediction of pro-
tein secondary structure and relative solvent accessibility using profiles,
machine learning and structural similarity, Bioinformatics. 30 (2014) 2592–
2597. https://doi.org/10.1093/bioinformatics/btu352.
19. S. Wang, J. Peng, J. Ma, J. Xu, Protein Secondary Structure Prediction Using
Deep Convolutional Neural Fields, Sci. Rep. 6 (2016). https://doi.org/10.1038/
srep18962.
20. Y. Ma, Y. Liu, J. Cheng, Protein secondary structure prediction based on data
partition and semi-random subspace method, Sci. Rep. 8 (2018). https://doi.
org/10.1038/s41598-018-28084-8.
21. Z. Aydin, Y. Altunbasak, M. Borodovsky, Protein secondary structure pre-
diction for a single-sequence using hidden semi-Markov models, BMC
Bioinformatics. 7 (2006). https://doi.org/10.1186/1471-2105-7-178.
22. R. Heffernan, K. Paliwal, J. Lyons, J. Singh, Y. Yang, Y. Zhou, Single-sequence-
based prediction of protein secondary structures and solvent accessibility
by deep whole-sequence learning, J. Comput. Chem. 39 (2018) 2210–2216.
https://doi.org/10.1002/jcc.25534.
23. A.K. Sharma, R. Srivastava, Protein Secondary Structure Prediction using
Character bi-gram Embedding and Bi-LSTM, (2019) 1–7.
24. A.K. Sharma, R. Srivastava, Variable Length Character N-Gram Embedding
of Protein Sequences for Secondary Structure Prediction, Protein Pept. Lett.
28 (2020) 501–507. https://doi.org/10.2174/0929866527666201103145635.
25. G. Wang, R.L. Dunbrack, PISCES: a protein sequence culling server,
Bioinforma. Appl. NOTE. 19 (2003) 1589–1591. https://doi.org/10.1093/
bioinformatics/btg224.
26. D. Wei, Q. Jiang, Y. Wei, S. Wang, A novel hierarchical clustering algo-

rithm for gene sequences, BMC Bioinformatics. 13 (2012) 174. https://doi.
org/10.1186/1471-2105-13-174.
27. R. Heffernan, Y. Yang, K. Paliwal, Y. Zhou, Capturing non-local interactions
by long short-term memory bidirectional recurrent neural networks for
improving prediction of protein secondary structure, backbone angles, con-
tact numbers and solvent accessibility, Bioinformatics. 33 (2017) 2842–2849.
https://doi.org/10.1093/bioinformatics/btx218.
28. S. Hochreiter, J. Schmidhuber, Long Short-Term Memory, Neural Comput.
(1997). https://doi.org/10.1162/neco.1997.9.8.1735.
29. A. Drozdetskiy, C. Cole, J. Procter, G.J. Barton, JPred4: A protein second-
ary structure prediction server, Nucleic Acids Res. (2015). https://doi.
org/10.1093/nar/gkv332.
30. S. Wang, W. Li, S. Liu, J. Xu, RaptorX-Property: a web server for protein
structure property prediction, Nucleic Acids Res. 44 (2016) W430–W435.
https://doi.org/10.1093/nar/gkw306.
8
Hesitancy, Awareness, and Vaccination:
A Computational Analysis
on Complex Networks
Dibyajyoti Mallick, Aniruddha Ray, Ankita Das and Sayantari Ghosh*
Department of Physics, National Institute of Technology Durgapur,

Durgapur, West Bengal, India
Abstract
The vaccines for preventing COVID-19 are being considered as the most effec-
tive way to reduce the high pandemic burden on the global health infrastructure.
However, public hesitancy towards vaccination is a crucial and pressing problem.
Our study has been designed to determine the parameters affecting the vaccina-
tion decisions of common individuals. Using the platforms of the compartmental
model and network simulation, we categorize people and observe their motivation
towards vaccinations in a mathematical social contagion process. We consider
peer influence as an important factor in this dynamic and study how individuals
influence their peers regarding vaccination decisions. The efficiency of the vacci-
nation process is estimated by the period of time required to vaccinate a substan-
tial fraction of the total population. We discovered the major barriers and drivers
of this dynamics and concluded that it is required to formulate specific strategies
targeting specifically the undecided and vaccine hesitant people.
Keywords: Complex network, mathematical model, social epidemic,

vaccination strategies
8.1 Introduction
The novel coronavirus disease 2019 (COVID-19), caused by severe acute
respiratory syndrome Coronavirus 2 (SARS-CoV-2), was first reported in
*Corresponding author: sayantari.ghosh@phy.nitdgp.ac.in
101
Wuhan, Hubei Province of China. This disease is now a global pandemic

which is spreading rapidly from person to person, causing major public
health concerns and economic crisis [1, 14]. A variety of active intervention
policies have been introduced to suppress the spreading of this disease such
as hand sanitizing, social distancing, travel restrictions, partial or complete
lockdown, wearing masks, quarantining, etc. After the declaration of the
pandemic by WHO (World Health Organization) in March 2020, phar-
maceutical companies and scientists have encountered a race against time
to develop vaccines [6]. The recent availability of multiple vaccines against
coronavirus has brought hope for prevention of the spreading and a rapid
recovery of our badly affected economy with a promise of a sooner resump-
tion of normal life. However, the widespread hesitancy about vaccines is
becoming a major obstacle for global health. Many people have strong hes-
itation towards vaccination, which is defined as the confusion about safety
and effectiveness of the vaccine. The origin of this usually lies in some
rumors regarding vaccines concerning side effects. This has become a huge
challenge for governments and public health authorities for reaching the
expected and required vaccination coverage [3, 9, 19]. The key priority now
is ensuring vaccine acceptance because the lag in vaccination may provide a
window for spreading the new variants and can also be a major obstacle in
developing society-wide herd immunity. Until now, many researchers and
scientists have tried to conduct different surveys to understand the behav-
ior towards vaccine acceptance and hesitancy [2, 8, 10]. Some studies [7]
have applied different theoretical models to explain vaccine acceptance,
hesitancy, and willingness of individuals towards vaccines as well as refusal
to vaccinate, which may vary depending upon personal decisions and epi-
demiological conditions [18]. Many advanced countries are trying to con-
duct country specific surveys [11], while several reports are being prepared
by different countries such as the United Kingdom (UK), United States of
America (USA), China, India, and Saudi Arabia to understand this vaccine
acceptance behaviour [5, 12, 16]. However, most of these studies are based
on heuristic arguments rather than mathematical analysis. A computational
framework having a mathematical foundation helps to draw quantitative
conclusions and is much more effective for predictive modeling purposes.
This paper represents a new mathematical model regarding the vaccina-
tion process which incorporates behavioral changes of every individual in a
society, driven by global and local factors. The main assumption here is that
the vaccination dynamics can be considered as a social contagion process.
This study aimed to identify the acceptability of the Covid-19 vaccine and
information in support or opposing vaccination that is flowing in a society
like a viral infection. We take into account refusal and hesitancy factors as
Hesitancy, Awareness, and Vaccination 103
well as the positive attitude of people towards vaccination [17]. In our study,
we focused on identifying the effects of the factors that could increase the
vaccination coverage through numerical simulations on an artificial society.
Moreover, we could justify several results found by survey based studies
where this kind of question has already been explored using survey results
[4, 15]. Our results address the vaccination acceptance problem using a
comprehensive mathematical and computational framework that would
help us figure out correct strategies to encourage community for vaccine
uptake and to stop further spreading of this pandemic.
8.2 Model Formulation

To understand this kind of problem, we have considered a set of differen-
tial equations to depict the possible transitions. To describe the vaccine
dynamics, we have a compartmental model, as shown in Figure 8.1. At any
time t, the total population N(t) is subdivided into four states: Ignorant (I),
Hesitant (H), Unwilling (U), and Vaccinated (V).
• Ignorant (I) − As the name suggests, in this group people are

ignorant about the vaccine. They have no idea about vacci-
nation and have no clear opinion. In general, these people
will not involve themselves in the vaccination process. But,
they may be influenced by hesitant and vaccinated groups as
well. Each person in this group is represented by I(t).
• Hesitant (H) − In this group people know about vaccines,
nevertheless they feel hesitation while making decisions
ΥUV -μU
µ α(H+V)I βH
I H U
- µl -μH λH
V
-μV
Figure 8.1 Compartmental diagram of proposed model (I-H-U-V).

All the transitions considered in the proposed model are shown through blue arrows. Sub-
populations are denoted by yellow boxes.
regarding vaccination, being affected by the rumors. This

group may be influenced by the global advertising cam-
paigns in favor of vaccination and take vaccines eventually.
On the other hand, they could also be influenced further by
rumors and transit to an unwilling state. Each person in this
group is represented by H(t).
• Unwilling (U) − In this group people believe in negative
rumors regarding vaccines and they push themselves from
vaccination. Again, some Unwilling people may return to the
Hesitant class if they cannot decide whether to get vaccine
or not, by influence through the vaccinated sub-population.
Each person in this group is represented by U(t).
• Vaccinated (V) – In this group, people are completely vacci-
nated and they are trying to influence other groups to join
them. Each person in this group is represented by V(t).
In this model, the key variable we have is the vaccinated population who
are influencing other populations to get vaccinated. This is also the target
population which has to be maximized over a certain period of time. Here,
this group of people is playing an effective role to control the epidemic and
spread the infodemic. Considering the total population as 1 (normalized
form),
I+H+U+V=1
Now, we will discuss the following possibilities of transition of

people from one sub-population to another over a given time period
‘t’. Let us discuss all the possible transitions considered in the model
categorically.
• Possible transitions related to I sub-population: In this pop-

ulation, µ is the rate at which people are entering into the
ignorant population. This is added to consider the demo-
graphic variations. Now, the members of I may be influ-
enced by the H and V population. H and V people might
spread words on their inclination and decision and encour-
age others for vaccination. Let us consider α to be the effec-
tive contact rate. If each member of I group is going to be
influenced by a hesitant and vaccinated group with that rate,
then α(H + V)I amount will be subtracted from the I group

and must be added to the hesitant group. In the normal
course, there must be some people who can die from that
population, hence the term µ should also be deducted from
group I in the rate equation.
• Possible transitions related to H sub-population: As men-
tioned in the previous case, due to the effect of influence
from group H and V, the term α(H + V)I should be added
in this rate equation. As usual, people from Group 3 can
also die, hence the term µH can also be deducted from H
group in its rate equation. Due to some reasons (influence
from other groups or due to development of self-awareness
among people), they are going for vaccination at a rate λ
and some are entering into the Unwilling group at a rate β.
Hence, the term βH and λH should be deducted from the
rate equation. Again, from the Unwilling group some peo-
ple are coming back into the Hesitant group by believing in
some rumors, hence the term γUV should be added to the
rate equation of H.
• Possible transitions related to U sub-population: From the
previous case discussed above it should be noted that the
terms γUV and µU should be deducted from the rate equa-
tion due to their hesitant behavior and natural death from
this group, respectively.
• Possible transitions related to V sub-population: From the
discussion of Case 2, it is evident that the term λH must be
added to its rate equation due to direct wishing for vaccina-
tion from group H and µV must be deducted due to natural
death of people from this group.
Hence, all the rate equations of this model are compiled as follows:
dI
= µ − α ( H + V )I − µ I (8.1)
dt
dH
= α ( H + V )I − β H − λ H + γ UV − µ H (8.2)
dt
dU
= β H − γ UV − µU (8.3)
dt
dV
= λ H − µV (8.4)
dt
8.3 Model Analysis on Complex Network

In ODE-based models, one of the major issues is homogeneous mixing,
which indicates every individual in a population has the same probability
of having contact with each other [20]. Our society is highly heterogeneous
and to accommodate that fact into our findings, we study the model on
the heterogeneous setting of a complex network. The simulations are per-
formed on a random network having 10000 nodes with an average degree
of 5. Here, we choose the EoN module in Networkx from python [13] and
run these following simulations in Google Colaboratory. Varying the tran-
sition rates of each parameter we see the effects of that particular parame-
ter on the time evolution of the dynamics through our model. The general
dynamics have been depicted in Figure 8.2. It shows that eventually most
of the people in the population get vaccinated, however the time of cover-
age might be different depending on the parameter values. Thus, to quan-
tify this growth curve, we define vaccination coverage time as ζV:
10000 VACCINATED
HESITANT
8000 UNWILLING
6000
4000
2000
0
0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0
Figure 8.2 Population of different classes vs. time.

Eventually most of the population gets vaccinated, but time of coverage depends on
parameters.
ζV = {t|V (t) = 0.9 Vmax}
This means that we consider a timescale, ζV, within which 90% of maxi-
mum vaccination, Vmax has been achieved. This will give us an estimate of
the vaccination coverage speed.
8.3.1 Effect of Vaccination Rate

Vaccination rate is a major parameter of this dynamic. If you care-
fully observe different sub-populations, the effect could be prominently
observed, as reported in Figure 8.3.
• Effect on vaccinated people: As we can see from the results

in Figure 8.3(a), increasing the rate does not change the
maximum number of vaccinated people, but it shifts the sat-
uration point leftwards which means the system is reaching
the saturation point faster.
• Effect on hesitant people: We can see from the plots shown
in Figure 8.3(b) that increasing the rate shifts the peak point
Total vaccinated population
10000
Total unwilling population
λ=0.2
3000 λ=0.5
λ=0.8
8000 2500
6000 2000
4000 1500
1000
2000 λ=0.2
λ=0.5 500
0 λ=0.8
0
0 5 10 15 20 25 30 35 40
0 5 10 15 20 25 30 35 40
time time
(a) (b)
24
Total hesistant population
4000 λ=0.2 22
3500 λ=0.5
λ=0.8 20
3000
2500 18
2000 16
1500 14
1000 12
500 10
0 8
0 5 10 15 20 25 30 35 40
time 0.2 0.3 0.4 0.5 0.6 0.7 0.8
(c) (d)
Figure 8.3 (a) Vaccination coverage with time for different values of vaccination rate λ.
(b) Time evolution of unwilling people in population with different values of vaccination
rate parameter. (c) Time evolution of hesitant people in population with different values
of vaccination rate parameter. (d) Vaccination coverage time (ζV) for different values of
vaccination rate.
slightly leftwards, which means the system is reaching the

peak point a bit faster and we also can see that the maximum
number of people reaching in a hesitant class decreases a lit-
tle bit.
• Effect on unwilling people: As we can see from the plots
shown in Figure 8.3(c), increasing the parameter flattens
the curve which means that the maximum number of peo-
ple going to the unwilling class decreases, which can be
explained by the success rate of the vaccination process.
As shown in Figure 8.3(d), we can see that the saturation time

decreases with increasing vaccination rate, that is λ, which states that
this particular parameter will make the system reach the saturation
point earlier.
8.3.2 Effect of Negative Rumors

Negative rumors often sway people towards some risky and dangerous
activity. In this vaccination process, some negative rumors about the vac-
cine exist that concern matters like several side effects, cost effectiveness,
fear, and misinformation regarding the vaccine. Hesitant people are often
influenced by them. The parameter that takes into account this phenomena
is β. The results related to this are shown in Figure 8.4.
• Effect on vaccinated class: As we can see in Figure 8.4(a),

increasing the parameter, β, shifts the saturation point of
vaccinated people rightwards. This means as more people
feel negative about vaccination, more time will be taken by
the system to reach the saturation point.
• Effect on unwilling class: In Figure 8.4(b), we can see that
many people are motivated by their neighbors, friends, or
relatives to believe some negative rumors about vaccination
and refuse to get vaccines and along the way the maximum
number of unwilling people increases.
8.3.3 Effect of Positive Peer Influence

Peer influence occurs when people are motivated or influenced towards
something by seeing their neighbor or friends. When someone’s peers
influence them to do something positive, it is considered as positive peer
influence. In the vaccination context, positive peer influence occurs when
Total vaccinated population
10000
2500

β = 0.2
β = 0.5
8000 β = 0.8
2000
6000 1500
4000 1000
2000 β = 0.2 500
β = 0.5
0 β = 0.8
0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time
(a) (b)
13.0
Total hesitant population
4000 β = 0.2
β = 0.5
β = 0.8 12.5
3000 12.0
11.5
2000 11.0
10.5
1000
10.0
0 9.5
0 5 10 15 20 25 30 35 40 9.0
time 0.2 0.3 0.4 0.5 0.6 0.7 0.8
(c) (d)
Figure 8.4 (a) Vaccination coverage with time for different values of negative rumor
influence rate β. (b) Time evolution of unwilling people in population with different
values of negative rumor influence rate parameter. (c) Time evolution of hesitant
people in population with different values of negative rumor influence rate parameter.
(d) Vaccination coverage time (ζV) for different values of negative rumor influence rate.
someone who is vaccinated is motivating others who are not sure about
vaccination. Here, the vaccinated people influence the hesitant people to
get vaccinated. Here we are considering two parameters α and γ that take
into account this phenomenon. The results related to this are shown in
Figure 8.5 and Figure 8.6.
• Effect on Vaccinated people: As we can see in Figure 8.5(a)

and 8.6(a), increasing the parameter shifts the saturation
point of vaccinated people leftwards. This means the more
people that feel positive about vaccination, then the sooner
the system will reach the saturation point.
• Effect on Unwilling people: We can also see from Figures
8.5(b) and 8.6(b) that along the way, the maximum number
of unwilling people decreases as they are getting motivated
by some positive rumors towards vaccination.
So, we can draw the conclusion that if more people are well informed
about the success rate of vaccination, then that would speed up the
Total vaccination population10000 Y=0.2

2500
Y=0.5
8000 Y=0.8
2000
6000 1500
4000 1000
2000 Y=0.2 500

Y=0.5
0 Y=0.8 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time
(a) (b)
3000 Y=0.2
Y=0.5
2500 Y=0.8 14
2000 13
1500
12
1000
11
500
0 10
0 5 10 15 20 25 30 35 40 0.2 0.3 0.4 0.5 0.6 0.7 0.8
time (d)
(c)
Figure 8.5 (a) Vaccination coverage with time for different values of positive peer
influence rate γ. (b) Time evolution of unwilling people in population with different
values of positive peer influence rate parameter. (c) Time evolution of hesitant people in
population with different values of positive peer influence rate parameter. (d) Vaccination
coverage time (ζV) for different values of positive peer influence rate.
10000
Total vaccination population
2000 α= 0.2
α= 0.5
8000 α= 0.8
1500
6000
1000
4000
2000 500
α= 0.2
α= 0.5
α= 0.8
0 0
0 5 10 15 20 25 30 35 40
time 0 5 10 15 20 25 30 35 40
time
(a) (b)
4000 α= 0.2 16
α= 0.5
α= 0.8 15
3000
14
2000 13
12
1000
11
0 10
0 5 10 15 20 25 30 35 40
time 0.2 0.3 0.4 0.5 0.6 0.7 0.8
(c) (d)
Figure 8.6 (a) Vaccination coverage with time for different values of positive peer
influence rate α. (b) Time evolution of unwilling people in population with different
values of positive peer influence rate α. (c) Time evolution of hesitant people in
population with different values of positive peer influence rate α. (d) Vaccination coverage
time (ζV) for different values of positive peer influence rate α.
vaccination process and also decrease the amount of unwilling people. As

shown in Figures 8.5(d) and 8.6(d), for both parameters of positive peer
influence, that is, α and γ, the saturation time decreases with increasing
parameters, which means positive peer influence will make the system
reach the saturation point earlier.
8.4 Conclusions and Perspectives

Along with the infectious spread of SARS-COV2, the information
and misinformation regarding available vaccines is also spreading
person-to-person, causing a large-scale effect on vaccination cover-
age. Through computational analysis and network simulations, we have
explored the contagion dynamics and proposed a framework to analyze
this decision process. From the model it has been observed that the pres-
ence of the efficient vaccination systems can run the entire dynamic with
positive feedback. This is a significant result, as we can see a lot of people
not only in India but around the world are hesitant about getting vacci-
nated. In our model, we have shown that if the peer influence is overall
positive, then the saturation point shifts leftwards, so we can conclude
that the spreading of awareness about the positive effects of vaccination
should be carried out on the ground level. We have also shown in our
model that if vaccination rate increases, then the saturation point of V
class shift leftwards. In real life we can explain it like this: if the vaccina-
tion rate in a particular area can be increased by awareness and smooth
execution of the vaccination process by a competent authority, then the
system would reach its saturation point sooner. A strong peer influence
also might signify a strongly connected society. The shift of the saturation
point leftwards for high positive peer influence means that physically and
virtually everyone is well connected. This may also indicate high vacci-
nation rates in urban areas. In the future we will focus on implementing
further realistic terms in our model. For example, if someone from a hes-
itant class goes to the unwilling class, it can be considered that the person
was under the influence of someone who has been vaccinated, thus we
can bring the non-linearity. Mathematical explorations are used to find
out the fixed points and bifurcations if there are any. Considering cou-
pled dynamics or delayed dynamics of disease along with the infodemic
might be another interesting future study.
References
1. W. Bank. The economy in the time of covid-19, 2020.
2. S. Bhattacharyya, A. Vutha, and C. T. Bauch. The impact of rare but severe
vaccine adverse events on behaviour-disease dynamics: a network model.
Scientific reports, 9(1):1–13, 2019.
3. A. A. Dror, N. Eisenbach, S. Taiber, N. G. Morozov, M. Mizrachi, A. Zigron, S.
Srouji, and E. Sela. Vaccine hesitancy: the next challenge in the fight against
covid-19. European journal of epidemiology, 35(8):775–779, 2020.
4. E. Dubé, M. Vivion, and N. E. MacDonald. Vaccine hesitancy, vaccine refusal
and the anti-vaccine movement: influence, impact and implications. Expert
review of vaccines, 14(1):99–117, 2015.
5. J. Eskola, P. Duclos, M. Schuster, N. E. MacDonald, et al. How to deal with
vaccine hesitancy? Vaccine, 33(34):4215–4217, 2015.
6. K. B. Habersaat and C. Jackson. Understanding vaccine acceptance
and demand—and ways to increase them. Bundesgesundheitsblatt-
Gesundheitsforschung-Gesundheitsschutz, 63(1):32–39, 2020.
7. C. Jarrett, R. Wilson, M. O’Leary, E. Eckersberger, H. J. Larson, et al. Strategies
for addressing vaccine hesitancy–a systematic review. Vaccine, 33(34):4180–
4190, 2015.
8. K. Kalimeri, M. G. Beiró, A. Urbinati, A. Bonanomi, A. Rosina, and C.
Cattuto. Human values and attitudes towards vaccination in social media. In
Companion Proceedings of The 2019 World Wide Web Conference, pages
248–254, 2019.
9. D. Kumar, R. Chandra, M. Mathur, S. Samdariya, and N. Kapoor. Vaccine
hesitancy: understanding better to address better. Israel journal of health
policy research, 5(1):1–8, 2016.
10. H. J. Larson, C. Jarrett, W. S. Schulz, M. Chaudhuri, Y. Zhou, E. Dube, M.
Schuster, N. E. MacDonald, R. Wilson, et al. Measuring vaccine hesitancy:
the development of a survey tool. Vaccine, 33(34):4165–4175, 2015.
11. J. V. Lazarus, S. C. Ratzan, A. Palayew, L. O. Gostin, H. J. Larson, K. Rabin, S.
Kimball, and A. El-Mohandes. A global survey of potential acceptance of a
covid-19 vaccine. Nature medicine, 27(2):225–228, 2021.
12. J. Luyten, L. Bruyneel, and A. J. van Hoek. Assessing vaccine hesitancy in
the UK population using a generalized vaccine hesitancy survey instrument.
Vaccine, 37(18):2494–2501, 2019.
13. J. C. Miller and T. Ting. Eon (epidemics on networks): a fast, flexible python
package for simulation, analytic approximation, and analysis of epidemics on
networks. arXiv preprint arXiv:2001.02436, 2020.
14. A. Nishi, G. Dewey, A. Endo, S. Neman, S. K. Iwamoto, M. Y. Ni, Y. Tsugawa,
G. Iosifidis, J. D. Smith, and S. D. Young. Network interventions for man-
aging the covid-19 pandemic and sustaining economy. Proceedings of the
National Academy of Sciences, 117(48):30285–30294, 2020.
15. P. Peretti-Watel, H. J. Larson, J. K. Ward, W. S. Schulz, and P. Verger. Vaccine

hesitancy: clarifying a theoretical framework for an ambiguous notion. PLoS
currents, 7, 2015.
16. P. Pronyk, A. Sugihantono, V. Sitohang, T. Moran, S. Kadandale, S. Muller,
C. Whetham, and R. Kezaala. Vaccine hesitancy in indonesia. The Lancet
Planetary Health, 3(3):e114–e115, 2019.
17. M. Sallam. Covid-19 vaccine hesitancy worldwide: a concise systematic
review of vaccine acceptance rates. Vaccines, 9(2):160, 2021.
18. M. Siddiqui, D. A. Salmon, and S. B. Omer. Epidemiology of vaccine hes-
itancy in the United states. Human vaccines & immunotherapeutics,
9(12):2643–2648, 2013.
19. G. Troiano and A. Nardi. Vaccine hesitancy in the era of covid-19. Public
Health, 2021.
20. N. Wang, Y. Fu, H. Zhang, and H. Shi. An evaluation of mathematical models
for the outbreak of covid-19. Precision Clinical Medicine, 3(2):85–93, 2020.
9
Propagation of Seismic Waves in Porous
Thermoelastic Semi-Infinite Space with
Impedance Boundary Conditions
Annu Rani and Dinesh Kumar Madan*
Department of Mathematics, Chaudhary Bansi Lal University, Bhiwani,

Haryana, India
Abstract
This work investigates the waves propagation in an isotropic saturated porous-
thermoelastic medium. The incidence of a seismic wave is considered at the free
surface of a thermoelastic semi-infinite space. The surface of semi-infinite space
is taken as isothermal. Reflection ratios of waves for the incidence of longitudi-
nal waves are derived for impedance boundary conditions. The graph depicts the
effect of porosity on reflection ratios versus incident angle and impedance param-
eter. Graph plotting is done with the help of the MATLAB programming language.
Results are presented for a particular model in this paper.
Keywords: Porous, thermoelastic, reflection, impedance, isothermal
9.1 Introduction
Thermoelasticity is the study of the change in shape and dimension of a
solid object when the temperature of the object varies. With the theory of
elasticity, Madan et al. [1] and Kumari and Madan [2] explore the propa-
gation of waves. By extending the classical coupled theory of thermoelas-
ticity by introducing relaxation time, Lord and Shulman [3] and Green
and Lindsay [4] presented the generalized theory of thermoelasticity. A
number of researchers [5–11] have discussed the phenomena of waves
*Corresponding author: dk_madaan@rediffmail.com
115
reflecting on a free surface. Rani and Madan [12] investigated the effect of
initial stress and imperfect interface on waves.
In porous media, the thermo-poroelasticity theory describes the rela-
tionship between temperature and stress components. A thermoelastic
saturated porous material with solid-fluid pores is a thermally conducting
porous solid and these materials are generally found in reservoir rocks of
the earth’s crust. The phenomenon of co-existence of thermoelasticity and
porosity plays a crucial role in the non-destructive evaluation of composite
materials and structures. The propagation of thermoelastic waves in fluid-
saturated soil was studied by Haibing [13]. Kumar et al. [14] investigated the
Rayleigh waves propagation in thermoelastic semi-infinite space with double
porosity under isothermal and insulated boundary conditions in thermo-
elastic semi-infinite space with double porosity. Wilson et al. [15] discuss the
propagation of acoustic waves. In a thermo-poroelastic medium, Wang et al.
[16] explored the reflection phenomenon of inhomogeneous elastic waves.
The reflection phenomenon of seismic waves at insulated and free surfaces of
thermo-poroelastic material was explored by Zorammuana and Singh [17].
Impedance boundary-conditions are defined as a linear combination
of unknown parameters and corresponding derivatives that are predict-
able on the surface. Electromagnetism, acoustics, seismology, science, and
technology all benefit from these conditions. The propagation of waves for
impedance boundary-conditions was studied by Singh [18], Vinh and Hue
[19], Vinh and Xuan [20], and others.
To conclude the wave propagation effects, a quantitative analysis of
reflection at the surface of a thermo-poroelastic medium is required.
This work examines the waves propagation at the surface of a saturated
porous-thermoelastic isothermal medium with an impedance boundary
condition. The influence of solid porosity and the Biot parameter for the
coupling of fluid and solid phases on the reflection ratios versus impedance
parameter and incident angle of a longitudinal wave is studied.
9.2 Basic Equations

The stress-tensors in a non-viscous thermal-porous saturated solid are
given by [21]:
σij = τij + β (−Pf) δij, (9.1)
where
τij: components of stress,
Pf: pressure of fluid,
Propagation of Seismic Waves 117
β: Biot parameter,
δij: Kronecker delta
The stress-strain relation is given by [22, 23]
τij = λekk δij + 2μeij − γs (ϴ − ϴ0) δij, (9.2)
Pf = −βNekk − Nvk, k + γs (ϴ − ϴ0) δij, (i, j, k = 1, 2, 3) (9.3)
where
λ, μ: lame’s constants,
eij: components of strain,
ϴ: temperature,
γs: coefficient of thermal stress corresponding to solid,
γf: coefficient of thermal stress corresponding to fluid,
N: elastic parameter for bulk-coupling of fluid,
vi = f (Ui − ui),
f: solid porosity,
Ui: component of displacement corresponding to fluid,
ui: component of displacement corresponding to solid,
1  ∂u ∂u 
Also, the strain components are emn =  m + n  , m, n = 1, 2, 3.
2  ∂xn ∂xm 
The equation of motion in an isotropic thermal-porous solid saturated
with a non-viscous fluid are given by [24]:
σ=
ij , j ρ ui + ρ f vi , (9.4)
(P=
f ), i ρ f ui + qvi , (9.5)
 ∂
Kθ,ii=  1 + t 0  ( ρ ceθ + γθ0 (u k , k + v k , k ) ) , (9.6)
 ∂t 
where
ρ: porous-aggregate density,
ρf: pore-fluid density,
q: coupling parameter between porous-aggregate solid and
pore-fluid,
γ = γs + βγf
9.3 Problem Formulation

Using Equations (9.1)-(9.3) in Equation (9.4), we have:
σ ij = λ ekkδ ij + 2 µeij − γ s (θ − θ 0 )δ ij + β ( β Nekk + Nv k ,k

− γ f (θ − θ 0 )δ ij ,(λ + µ + β 2 N )uk ,ki + µuk ,ii − γθ ,i − ρvk
(9.7)
+ β Nv k ,ki − ρ f vk = 0
β Nuk , ki + Nvk , ki − γ f θ ,i −qvk − ρ f uk = 0 (9.8)
 ∂
Kθ,ii=  1 + t 0  ( ρ ceθ + γθ0 (u k , k + v k , k ) ) (9.9)
 ∂t 
Displacement components in terms of potentials are
∂ps ∂qs ∂p f ∂q f
u1 = − , u3 = + (9.10)
∂x ∂z ∂z ∂x
∂p f ∂q f ∂p f ∂q f
v1 = − , v3 = + (9.11)
∂x ∂z ∂z ∂x
Using Equations (9.10) and (9.11) in Equations (9.7)-(9.9), we have:
(λ + µ + β 2 N )∇ 2 ps + µ∇ 2 ps − γθ − ρ ps + β N ∇ 2 p f − ρ f p f = 0 (9.12)
µ∇ 2qs − ρqs − ρ f q f = 0 (9.13)
β N ∇2 ps − γ f θ − qp f + N ∇2 p f − ρ f ps = 0 (9.14)

qq f + ρ f qs =
0 (9.15)
ρ ce (θ + t 0θ) + γθ0 (∇2 p s + t 0∇2 ps + ∇2 p f + t 0∇2 p f ) − K ∇2θ = 0 (9.16)
∂2 ∂2
where ∇2 = 2 + 2 is the Laplace operator.
∂x ∂z
The coupling structure of potentials pf , ps with thermal wave can be eas-
ily observed from Equations (9.12), (9.14), and (9.16), while the coupling
structure of potentials qf and qs can be easily observed from Equations
(9.13) and (9.15).
The velocities of waves are [17]
1
 H1 H 2 − H 3 3
 6 2 
 1 
  2 3
2  H H12 
V1 =   H 2 − H1  +   − 
2
−
  3 9   H13   3 9 
+  3 2
−
27 
   H 1 H H
1 2 H 3   
− +
  27 6 2 
  
1
 H1 H 2 − H 3 3
 6 2 
 1 
  2 3
2  H
  H 2 − H1  +   − 1 , (9.17)
  3 9   H1  3 3
+  3 2
−
27 
  H 1 − H1 H 2 + H 3   
  27 6 2 
  
( )
2
H2 H1
−
H1
3 9
V2 = −
 ( )  
1
 H H
{( ) ( )}
1
3
3 2
2 
2 3 2 3
H3 H2 H1 H1 H1 H2 H3 H1
 
1 2
− + − + − + −
6 2 3 9 27 6 2 27
1
{( ) ( )}
 
1
3
3 2
H H 
2 3 2 3
1 2
H3 H2 H1 H1 H1 H2 H3 H1
 6 −
2
+
3
−
9 27
−
6
+
2
−
27

−
2

{ } 
1
1 2
1
{( ) ( )}
 3
2
H2
−
H1
 3 2
2
 3
H H 
2 3 3
H3 H2 H1 H1 H1 H2 H3 H
 3 9
+
 6
1 2
− + − + − + −
1

 
1
2 3 9 27 6 2 27
H H
{( ) ( )}

1

3
3 2
 
2 3 2 3
H3 H2 H1 H1 H1 H2 H3 H1
1 2
− + − + − + −
 i
6 2 3 9 27 6 2 27
+
2
(9.18)
2
H2 H1
−
3 9 H1
V3 = 1
−
3
   1
3
2  H H H3  H H1
2

3
H 3
H1H 2 H3   H 
2
2 3
+   +   − 
1 2 2 1 1
− − − +
  6 2  3 9   27 6 2   27 
1
  1
3
H H H3  H H1
2

3
H 3
H1H 2 H3   H 
2
2 3
+   +  − 
1 2 2 1 1
− − − +
 6 2  3 9   27 6 2   27 
− −
2

{ } 
1 2
H2 H1
 32 − 
3 9
 1
+
 1
 
  H H
3
H3  H H1
2

3
H 3
H1H 2 H3 
2
2 3
H1  
 +   +  i
1 2 2 1
− − − + −
  27 
 6 2  3 9 6 2  27  
 1

  
1
 H
3

  H 
3 2
 H H H3 H1
2
 H 3
H1H 2 H3
2 3

+   +  − 
1 2 2 1 1
− − − +
  6 2  3 9   27 6 2   27  
(9.19)
2
µ
Vs = , (9.20)
ρ 2f
ρ−
q
η1 η2 η3 i
=
where, H1 = , H2 ,=
H3 = , t t0 + ,
η0 η0 η0 ω
η1 = ρCet (Dq + N ρ − 2 β N ρ f ) + K ( ρq − ρ 2f ) + θ 0tγγ f ( ρ − ρ f )

+γ 2θ 0t (q − ρ f ),
η2 = − K (Dq + ρ N − 2β N ρ f ) − ρtCe (DN − β 2 N 2 ) − θ 0tγγ f (D − β N )
+θ 0tγ 2 ( β N − N ),
η3 = (λ + 2µ )KN ,η0 = ρCet ( ρ 2f − qρ ) (9.21)

9.4 Reflection at the Free Surface

Consider an isotropic porous-thermoelastic semi-infinite space, z > 0
(Figure 9.1). If a longitudinal wave is incident with an angle e at the surface,
then it will give four reflected waves. Let e1, e2, e3 be angles which reflect
coupled longitudinal waves, making them normal, and es be the angle
which reflected transverse waves makes normal in medium M.
x axis
Isotropic porous e1
thermoelastic semi- e e2
D1
infinite space (M) e3
es C3
C0 C2
C1
z axis
Figure 9.1 Geometry of problem for reflection of waves when longitudinal wave is
incident.
=ps C0e i (k0 ( xsin(e ) + zcos (e )) −ωt ) + C1e i (k1 ( xsin(e1 ) − zcos(e1 )) −ωt )
+C2e i (k2 ( xsin(e2 ) − zcos (e2 )) −ωt ) + C3e i (k3 ( xsin(e3 ) − zcos(e3 )) −ωt ) , (9.22)
=p f a0C0e i (k0 ( xsin(e ) + zcos (e )) −ωt ) + a1C1e i (k1 ( xsin(e1 ) − zcos(e1 )) −ωt )
+a C e 2 (
i ( k xsin ( e2 ) − zcos ( e2 ) ) −ωt )
2 2 + a C e i (k3 ( xsin(e3 ) − zcos (e3 )) −ωt ) (9.23)
3 3
qs = D1e i (ks ( xsin(es ) − zcos (es )) −ωt ) , (9.24)
q f = d1D1e i (ks ( xsin(es ) − zcos (es )) −ωt ) , (9.25)
b0C0e i (k0 ( xsin(e ) + zcos (e )) −ωt ) + b1C1e i (k1 ( xsin(e1 ) + zcos (e1 )) −ωt )
θ − θ0 =
+b2C2e i (k2 ( xsin(e2 ) − zcos (e2 )) −ωt ) + b3C3e i (k3 )( xsin(e3 ) − zcos (e3 )) −ωt ) , (9.26)
where
( D − ρV02 ) γγ f − ( β N − ρ f V02 )
a0 = ,
γf
( N − qV02 ) − ( β N − ρ f V02 ) γ
b =
0
(( N − qV )( D − ρV ) − ( β N − ρ V ) ) k
0
2
0
2
f 0
2 2 2
0
,
 2 γ f 2 
 ( β N − ρ f V0 ) γ − ( N − qV0 ) γ
( D − ρVi2 ) γγ f − ( β N − ρ f Vi2 )
ai = , (9.27)
( N − qVi ) − ( β N − ρ f Vi ) γγ f
2 2
b =
i
(( N − qV )( D − ρV ) − ( β N − ρ V ) ) k
i
2
i
2
f i
2 2
i
2
, i = 1,2,3,
 2 γ f
 ( β N − ρ f Vi ) γ − ( N − qVi2 ) γ
µ − ρVs2 2
d1 = 2 , D = λ + 2µ + β N
ρ f Vs
Using Snell’s Law, angles and corresponding wave vectors are associated
by:
k1 sin (e) = k1 sin (e1) = k2 sin (e2) = k3 sin (e3) = ks sin (es) (9.28)
and in terms of velocities, are given by:
sin(e) sin(e1 ) sin(e2 ) sin(e3 ) sin(es )

= = = = (9.29)
V0 V1 V2 V3 Vs
9.4.1 Boundary Conditions

At surface z = 0, impedance boundary conditions are given by [18]:
1. σ33 = 0, which gives:
 2 2 γ b1  k12
 ( λ + β N ) + 2 µ cos e + a β N +
k12  k02
1 1
 γ b  k2
X1 +  (λ + β 2 N ) + 2 µ cos 2 e2 + a2 β N + 22  22
 k2  k0
 γ b  k2 k2
X 2 +  (λ + β 2 N ) + 2 µ cos 2 e3 + a3 β N + 23  32 X 3 − 2 µ sin es cos es s2
 k3  k0 k0
 γb 
X 4 =  (λ + β 2 N ) + 2 µ cos 2 e + a0 β N + 20 
 k0 
(9.30)
2. σ31 + ωZu1 = 0 gives:
 2k12 k1   2k22 k 
 2 sin e1 cos e1 − iω Z 2
sin e X +
1 1  2
sin e2 cos e2 − iω Z 22 sin e2 
 k0 k0   k0 k0 
 2k 2 k 
X2 +  23 sin e3 cos e3 − iω Z 32 sin e3 
 k0 k0 
 k 2
k 
X3 +  (cos2 es − sin2 es ) s2 − iω Z 2s cos es 
 k0 k0 
iω Z
X 4 = 2 sin e0 cos e0 + sin e0
k0
(9.31)
3. v 3 = 0 gives:
k1 k k k
a1 cos e1 X1 + 2 a2 cos e2 X2 + 3 a3 cos e3 X3 − s d sin es
k0 k0 k0 k0
(9.32)
X 4 = a0 cos e,
∂θ
4. + hθ =0 (h → 0 for thermally-insulated surface, h → ∞
∂y
for isothermal surface) gives
−ik1 a1 X1 cos (e1) − ik2 a2 X2 cos (e2) − ik3 a3 X3 cos (e3)
+ h (b1 X1 + b2 X2 + b3 X3) = −b0h + ik0 cos e. (9.33)
where
C1 C2 C3 D1
=X1 = , X2 = , X3 = , X4 , (9.34)
C0 C0 C0 C0
The matrix equation from these equations can be written as
BX = Y, (9.35)
where
b11 b12 b13 b14

  X1   z1 
b  X   
B=
21 b22 b23  , X =  2  ,Y =  z 2 
b24
b31 b32 b33 
b34  X3   z3 
     
b41 b42 b43 b44
  X 4   z 4 
 2 2 γ b1  k12
b=  (λ + β N ) + 2 µ cos e + a β N + ,
k12  k02
11 1 1

 2 2 γ b2  k22
b=  ( λ + β N ) + 2 µ cos e + a β N + ,
k22  k02
12 2 2

 2 2 γ b3  k32
b=  (λ + β N ) + 2 µ cos e + a β N + ,
k32  k02
13 3 3

k2  2k 2 k 
b14 = −2 µ sin es cos es s2 , b21 =  21 sin n e1 cos e1 − iω Z 12 sin e1  ,
k0  k0 k0 
 2k22 k 
=b22  2 sin e2 cos e2 − iω Z 22 sin e2  ,
 k0 k0 
 2k 2 k 
b23 =  23 sin e3 cos e3 − iω Z 32 sin e3  ,
 k0 k0 
 k2 k 
b24 =  (cos 2 es − sin 2 es ) s2 − iω Z s2 cos es  ,
 k0 k0 
k1 k k
b31 = a1 cos e1 , b32 = 2 a2 cos e2 , b33 = 3 a3 cos e3 ,
k0 k0 k0
k
b34 = − s d sin es , b41 = ( −ik1a1 cos(e1 ) + hb1 ),
k0
b42 = ( −ik2a2 cos(e2 ) + hb2 ), b43 = ( −ik3a3 cos(e3 ) + hb3 ),
γb
b44 = 0, z1 =  (λ + β 2 N ) + 2µ cos 2 e + a0 β N + 20  ,
 k0 
iω Z
z 2 = 2sin e0 cos e0 + sin e0 , z3 = a0 cos e ,
k0
z 4 = −b0h + ik0 cos e.
(9.36)
Here, Xi, i = 1, 2, 3, 4 represents reflection ratios in isotropic thermoelas-

tic porous medium.
9.4.2 Energy Ratios

The average energy transmission per unit surface area per unit time is
given by [25]:
=P σ 33u 3 + σ 31u1 − Pf β v 3 (9.37)
Using Equations (9.22)-(9.28) and (9.37),
PrefL1 R1 2 PrefL 2 R2 2
E1 = = X1 , E 2 = = X2 ,
Pinc L1 R0 Pinc L 2 R0
PrefL 3 R3 2 PrefT R4 2
E3 = = X 3 , E4 = = X4 , (9.38)
Pinc L 3 R0 Pinc T R0
where PrefL1, PrefL2, PrefL3, and PrefT are the average energy of reflected coupled
longitudinal waves at e1, e2, e3, and transverse wave at e4, respectively, and
 γ b γ f b0a0 β  3
R0 = −  λ + 2 µ + β 2a0 N + a02 β N + s 2 0 + k0 cos e ,
 k0 k02 
 γ b γ f b1a1 β  3
R1 =  λ + 2 µ + β 2a1 N + a12 β N + s 2 1 + k1 cos e1 ,
 k1 k12 
 γ b γ f b2a2 β  3
 k2 k22 
 γ b γ f b3a3 β  3
 k3 k32 
R4 = µ ks3 cos es . (9.39)
9.5 Numerical Results and Discussion

For the numerical computation of amplitude ratios, we employed the fol-
lowing parameters [24]:
J W
λ = 3.7 GPa, µ = 7.9 GPa, N = 6GPa, Ce = 1040 , K = 170 ,
kg K m
kg kg ρf
ρ = 2216 3 , ρ f = 950 3 , q = 1.05 ,
m m f
GPa
β f = 2.37 × 10 −3 , β s = 2 β f ,ω = 2s −1 ,
K
T0 = 300° K , t0 = 10 −10 s
The problem of isothermal surfaces has been studied here. To plot the
graphs and execute numerical calculations, the MATLAB programming
language is employed. Figure 9.2 explored the impact of porous-parameter
f on the reflection ratio of a coupled longitudinal wave at an angle e1 to the
incident angle e of the longitudinal wave. It is concluded that as the value of
f increases, the reflection ratio of a longitudinal wave decreases for e < 70°
1.05
0.95
0.9
0.85
X1
0.8
0.75
0.7
0.65
0.6
0 10 20 30 40 50 60 70 80 90
e (in degrees)
Figure 9.2 Reflection ratio X1 of longitudinal wave (at angle e1) against e (incident angle)
for β = 0.04, Z = 10, f = 0.03 (circle-line), f = 0.04 (star-line), f = 0.05 (triangle-line).
and then for e > 70°, attaining the same value for f = 0.03 (circle-line), f =
0.04 (star-line), and f = 0.05 (triangle-line).
Variation of the reflection ratios of coupled longitudinal waves at an
angle e2, e3 against the incident angle for β = 0.4, Z = 10, f = 0.03 (circle-line),
f = 0.04 (star-line), f = 0.05 (triangle-line) is shown in Figures 9.3 and 9.4,
respectively. It is concluded that the reflection ratio of a coupled-longitu-
dinal wave against the incident angle attains minimum value at e = 71°,
then e > 71° attains the same value for distinct values of f. Qualitatively, the
×10–8
9
5
X2
4
0
0 10 20 30 40 50 60 70 80 90
e (in degrees)
Figure 9.3 Reflection ratio X2 of longitudinal wave (at angle e2) against incident angle e
0.045
0.04
0.035
0.03
0.025
X3
0.02
0.015
0.01
0.005
0
0 10 20 30 40 50 60 70 80 90
e (in degrees)
Figure 9.4 Reflection ratio X3 of longitudinal wave (at angle e3) against incident angle e
same behavior of the reflection ratios of a longitudinal wave at an angle e2,

e3 against the incident angle is observed.
Figure 9.5 explores the variation in amplitude ratio of transverse wave
against incident angle e of the longitudinal wave. The reflection ratio of
a coupled longitudinal wave attains a maximum value at e = 8° and then
decreases gradually. The reflection ratio of a transverse wave behaves qual-
itatively the same as a coupled longitudinal wave at an angle e2, e3.
Figures 9.6 and 9.7 exhibit variation in the amplitude ratio of a cou-
pled longitudinal wave at angle e1, e2 against impedance parameter Z
for f = 0.01, β = 0.1 (represented by circle-line), β = 0.2 (represented by
triangle-line), and β = 0.3 (represented by star-line). The magnitude of
4
3.5
3
2.5
X4
2
1.5
1
0.5
0
0 10 20 30 40 50 60 70 80 90
e (in degrees)
Figure 9.5 Reflection ratio X4 of transverse wave against incident angle e (in degrees) for
β = 0.04, Z = 10, f = 0.03 (circle-line), f = 0.04 (star-line), f = 0.05 (triangle-line).
1.08
1.075
1.07
1.065
X1
1.06
1.055
1.05
1.045
1.04
0 5 10 15 20 25 30 35 40 45 50
Z
Figure 9.6 Reflection ratio X1 of longitudinal wave at e1 against Z (impedance parameter)

for f = 0.01, e = 30°, β = 0.1 (circle-line), β = 0.2 (triangle-line), β = 0.3 (star-line).
×10-8
6
4
X2
3
0
0 5 10 15 20 25 30 35 40 45 50
Z
Figure 9.7 Reflection ratio X2 of longitudinal wave at e2 against Z (impedance parameter)

amplitude ratios of coupled longitudinal wave at angle e1, e2 increases

as the value of β increases and attains a maximum value at Z = 50. The
amplitude ratios of a coupled longitudinal wave at an angle e3 and trans-
verse wave increases against the impedance parameter Z and attains a
maximum value at Z = 50 in Figures 9.8 and 9.9, respectively. It is noticed
that the quantitative value of the reflection ratio decreases as the value of
β increases.
0.014
0.013
0.012
0.011
X3
0.01
0.009
0.008
0.007
0 5 10 15 20 25 30 35 40 45 50
Z
Figure 9.8 Reflection ratio X3 of longitudinal wave at e3 against Z (impedance parameters)

0.5
0.45
0.4
0.35
X4
0.3
0.25
0.2
0.15
0.1
0 5 10 15 20 25 30 35 40 45 50
Z
Figure 9.9 Reflection ratio X4 of transverse wave against Z (impedance parameter) for
f = 0.01, e = 30°, β = 0.1 (circle-line), β = 0.2 (triangle-line), β = 0.3 (star-line).
Effect of Biot parameter on energy ratios of transverse and longitudi-

nal waves against incident angle of longitudinal wave is shown in Figures
9.10-9.13. It is noticed that the energy ratios of longitudinal and transverse
waves achieve a fixed value for incident angle e > 70°. As the value of the
Biot parameter grows, the energy ratio of longitudinal waves reflected at
an angle e1 decreases, while the energy ratio of longitudinal and transverse
waves reflected at angles e1, e2 and e3, respectively, increases against the
incident angle of longitudinal waves.
1.1
0.9
0.8
0.7
E1
0.6
0.5
0.4
0.3
0 10 20 30 40 50 60 70 80 90
e (in degrees)
Figure 9.10 Energy ratio E1 of longitudinal wave at e1 against incident angle e for f = 0.05,
Z = 10, β = 0.3 (circle-line), β = 0.4 (star-line), β = 0.5 (triangle-line).
×10-8
4.5
4
3.5
3
2.5
E2
2
1.5
1
0.5
0
0 10 20 30 40 50 60 70 80 90
e (in degrees)
0.5
0.45
0.4
0.35
0.3
0.25
E3
0.2
0.15
0.1
0.05
0
0 10 20 30 40 50 60 70 80 90
e (in degrees)
×10-3
3.5
2.5
2
E4
1.5
0.5
0
0 10 20 30 40 50 60 70 80 90
e (in degrees)
Figure 9.13 Energy ratio E4 of transverse wave against incident angle e for f = 0.05,
9.6 Conclusion
The effects of porosity and impedance on waves propagating in isotropic-
porous thermoelastic semi-infinite space is investigated in this work. The
following key points are drawn from graphs in this paper:
• As solid porosity increases, the amplitude ratio of the longi-

tudinal wave reflected at an angle e1, drops against the inci-
dent angle of longitudinal waves, reaching a minimum value
at e = 65° and subsequently, a fixed value for e > 70°
• The reflection ratios of longitudinal waves (at angle e2, e3)
and transverse wave falls against incident angle of longitu-
dinal wave, as solid porosity increases, reaching a minimum
for e > 70°
• Amplitude ratios of reflected waves in isotropic porous-
thermoelastic medium increases with the impedance parameter
• The reflection ratio of the coupled longitudinal waves at
angle e1, e2, increases against the impedance parameter with
increment of Biot parameter
• The reflection ratio of the longitudinal wave at e3 and trans-
verse wave decreases against the impedance parameter with
increment of Biot parameter
• The Biot parameter has less impact on the reflection ratio of
the transverse waves
• The energy ratio of the reflected longitudinal waves at an
angle e1 drops against the incident angle of longitudinal
waves, achieving a minimum value at e = 65° and subse-

quently a fixed value for e > 70°
• The sum of all energy ratios approaches unity
References
1. D.K. Madan, A. Rani and M. Punia, A note on the effect of rigidity and initial
stress on the propagation of Rayleigh waves in pre-stressed orthotropic elas-
tic layered medium, Proc. Indian Natl. Sci. Acad., Vol. 87, pp. 487–498, 2021.
2. A. Kumari and D.K. Madan, Deformation field due to seismic sources with
imperfect Interface, J. Earth Syst. Sci., Vol. 130, pp. 1-19, 2021.
3. H. Lord and Y. Shulman, A generalized dynamical theory of thermoelastic-
ity, J. Mech. Phys. Solids, Vol. 15, pp. 299-309, 1967.
4. A.E. Green and A. Lindsay, Thermoelasticity, J. Elast., Vol. 2, pp. 1-7, 1972.
5. B. Singh, Reflection of P and SV waves from free surface of an elastic solid
with generalized thermodiffusion, J. Earth Syst. Sci., Vol. 114, pp. 159-168,
2005.
6. A. N. Abd-Alla and A.A.S. Al-Dawy, The reflection phenomenon of SV waves
in a generalized thermoelastic medium, Int. J. Math. Sci., Vol. 23, pp. 529-
546, 2000.
7. J. N. Sharma, V. Kumar and D. Chand, Reflection of generalized thermo-
elastic waves from the boundary of a half-space, J. Thermal Stresses, Vol. 26,
pp. 925-942, 2003.
8. S. Abo-dahab, R.A. Mohamed and B. Singh, Rotation and magnetic field
effects on P wave reflection from a stress–free surface of elastic half-space
with voids under one thermal relaxation time, Journal of Vibration and
Control, Vol. 17, pp. 1827-1839, 2010.
9. R. Bijarnia and B. Singh, Propagation of plane waves in an anisotropic gen-
eralized thermoelastic solid with diffusion, Journal of Engineering Physics
and Thermophysics, Vol. 85, pp. 442-448, 2012.
10. S. B. Sinha and K. A. Elsibai, Reflection of thermoelastic waves at a solid half-
space with two thermal relaxation times, Journal of Thermal Stresses, Vol. 20,
pp. 129-146, 1996.
11. A.N. Sinha and S.B. Sinha, Reflection of thermoelastic waves at a solid
half-space with thermal relaxation, Journal of Physics of the Earth, Vol. 22,
pp. 237-244, 1974.
12. A. Rani and D.K. Madan, Effect of initial stress and imperfect interface on
love waves propagation in prestressed orthotropic layer coated over a pre-
stressed orthotropic semi-infinite space, Journal of Rajasthan Academy of
Physical Sciences, Vol. 20, pp. 219-228, 2021.
13. T. Haibing, L. Ganbin, X. Kanghe, Z. Rongyue and D. Yuebao, Characteristics

of wave propagation in the saturated thermoelastic porous medium, Transp
Porous Med., Vol. 103, pp. 47-68, 2014.
14. R. Kumar, R. Vohra and S. M. Abo-Dahab, Rayleigh waves in thermoelastic
medium with double porosity, MOJ Civil engineering. Vol. 4, pp. 143-148,
2018.
15. R.K. Wilson and E. S. Aifantis, A double porosity model for acoustic wave
propagation in fractured porous rock, Int. J. Engg. Sci., Vol. 22, pp. 1209-
1227, 1984.
16. E. Wang, J.M. Carcoine, Y. Yuan and J. Ba, Reflection of inhomogeneous
plane waves at the surface of a thermo-poroelastic medium, Geophysical
Journal International, Vol. 224, pp. 1621-1639, 2021.
17. C. Zorammuana and S. S. Singh, Elastic waves in thermoelastic saturated
porous Medium, Meccanica, Vol. 51, pp. 593-609, 2016.
18. B. Singh, Reflection of plane waves from surface of a generalized thermo-
viscoelastic porous solid half-space with impedance boundary conditions,
Mechanics and Mechanical Engineering, Vol. 22, pp. 1483-1496, 2018.
19. P.C. Vinh and T.T. Hue, Rayleigh waves with impedance boundary condi-
tions in anisotropic solids, Wave Motion, Vol. 51, pp. 1082-1092, 2014.
20. P.C. Vinh and N.Q. Xuan, Rayleigh waves with impedance boundary con-
dition: Formula for the velocity, existence and uniqueness, Eur. J. Mech. A
Solids, Vol. 61, pp. 180-185, 2017.
21. M. A. Biot, Mechanics of deformation and acoustic propagation in porous-
media, J Appl Phys, Vol. 33, pp. 1482-1498, 1962.
22. J. Bear, S. Sorek, G. Ben-Dor and G. Mazor, Displacement waves in satu-
rated thermoelastic porous media, I. Basic Equations. Fluid Dyn Res, Vol. 9,
pp. 155-164, 1992.
23. A. Levy, S. Sorek, G. Ben-Dor and J. Bear, Evolution of the balance equations
in saturated thermoelastic porous media following abrupt simultaneous
changes in pressure and temperature, Trans Porous Media, Vol. 21, pp. 241-
268, 1995.
24. M.D. Sharma, Wave propagation in thermoelastic saturated porous medium,
J Earth Syst Sci., Vol. 117, pp. 951-958, 2008.
25. J.D. Achenbach, Wave propagation in elastic solids, North-Holland, Amsterdam,
1973.
10
IoT Based Ensemble Predictive
Techniques to Determine the Student
Observing Analysis through E-Learning
Rufia Thaseen I.1, S. Shahar Banu1* and Sudha Rajesh2
1
Department of Computer Applications, B. S. Abdur Rahman Crescent Institute of
Science & Technology, Vandalur, Tamil Nādu, India
2
Department of Computational Intelligence, SRMIST, Kattankulathur,
Tamil Nādu, India
Abstract
The objective was to make a comparative study of the IoT based Ensemble
Predictive system with real-life teacher predictions on student observation
during E-Learning. Data is collected from 46 faculties for 188 periods through an
opinion-based survey using a questionnaire. Similarly, for the 188 periods the data
was collected from the created IoT based Ensemble Predictive System. The system
is designed in such a way that it uses five variables, namely: Level of Interaction,
No. of Questions Raised, No. of Students in the Class, No. of Concepts Taught in
a Period, and Responsiveness of the Students to Questions during Class Hours, to
perform the student observation analysis. From observation, it is interpreted that
there is no major difference in the solution provided by the faculties and output
generated by the IoT Ensemble Predictive system. Further, it was found that there
is a remarkable relationship between the opinion provided by the faculties and
output generated by the IoT Ensemble Predictive system. But for the item (No.
of Students in the Class), there is no significant relationship between the opinion
provided by the faculties and output generated by the IoT Ensemble Predictive sys-
tem. Therefore, it can be interpreted that when the number of students increased
or decreased beyond ideal conditions, there is the possibility of deviation in the
opinion provided by the faculties and output generated by the system. Also, the
calculated R-value is 0.788, meaning there is a 78.8% strong positive relationship
between the opinion provided by the faculties and output generated by the system.
*Corresponding author: shahar@crescent.education
137
The estimated R-Square Value (0.621) indicates that it is possible to forecast an

IoT based Ensemble Predictive system in the real world with 62.1% efficiency.
Furthermore, the evaluated.
Coefficients’ significance value is not more than 0.05 and this confirms that an
IoT based Ensemble Predictive system can be forecasted with real-world informa-
tion and vice-versa.
Keywords: IoT, education system, E-learning
10.1 Introduction
The World Wide Web was very efficient in terms of posting and retrieving
data in the beginning stages. The World Wide Web was only a collection of
static web pages with links to other sites and users could access the infor-
mation they needed by surfing and moving from one page to the next
[1]. After some time, the situation changed and web 2.0 was born. A user
of a website could do significantly more with it than they could with a
web 1.0 site, such as reading material and navigating through one page to
another page through hyperlinks. The normal person’s engagement with
online resources is easily available. Social networking sites like Facebook,
YouTube, and Flickr, among others, are examples of web 2.0 applications.
We are currently living in the age of semantic web. Computers in web 3.0
are capable of intelligently evaluating information available on web sites, as
well as developing and propagating the information they contain [2].
Until recently, the internet was designed to bring people together, enabling
them to communicate, exchange information, send messages, and hold vir-
tual conferences. We now have machine to machine interactions in addition
to human to human and human to machine interactions thanks to the intro-
duction of web 3.0. This is a huge step forward. System-to-system communi-
cation can be shown through a communication linking a temperature sensor
and AC [3]. Consider the following scenario: a temperature sensor monitors
the room temperature and when it rises beyond a certain threshold, it sends
an indication to the AC prompting it to immediately turn on the switch. If
the temperature goes down a specific point [4], the temperature sensor will
send a second signal to the air conditioner, this time instructing it to turn off
the machine. M2M is a technology that is used in the IOT. With the growth
of technology, it is now feasible for people to engage with and contribute
information on a website in a conventional manner. The information that
can be utilized in the reception and transmission of analyzed data may be
used to its full capacity by the specific kind of connection of different things
that can be achieved via the internet [5].
IoT Based Ensemble Predictive Techniques 139
As examples, a lot of web 2.0 apps are available. YouTube, Flickr,

Facebook, and a slew of other sites are among them. Consumers are
increasingly familiar with web 3.0, which is a semantic web that enables
them to explore with more sophistication. Because of the existence of
web 3.0, persons who use computers can understand the availability of
information in a more intellectual manner [6]. Web 3.0 can both send
and develop information depending on the data it receives from its data
sources. The internet may be considered one of the available resources
for individuals to interact and communicate with one another, as well
as for the spread and acquisition of relevant information. It may also be
used for information transfer and in the sophisticated version, video con-
ferencing can be utilized to make information accessible. Apart from the
advancement of man-to-man communication, the promotion of web 3.0
may also be beneficial to the advancement of man-to-machine commu-
nication. Web 3.0 also allows for machine-to-machine communications
[7]. Every day, more smart items that are tailored to certain situations are
produced and they are becoming increasingly common in many fields of
education, including higher education. Bright education is the most sig-
nificant component in establishing IoT bright cities that is made possible
by virtual learning and digital transformation when it comes to smart ser-
vices that are backed by developing technologies such as the IoT. Applying
IoT in online learning, when utilized in bright cities, allows e-citizens to
be connected and imaginative with a higher degree of involvement and
teamwork in the learning activity as well as other decisions made by their
peers and the community [8].
The implementation of IoT in smart cities has a great impact on online
learning processes by offering a comprehensive electronic approach to the
online educational community’s resources through a centrally integrated
system in a smart fashion. Academic work shows the ability and effect of the
IoT in smart city e-education and e-learning processes are expected, mea-
surable, and cannot be overlooked. Hyper-connection linking items, higher
degree of access, scalability, and integration of communication networks
(RFID and WSNs) are Internet of Things features that might improve the
effectiveness of e-learning techniques in “smart settings,” including brighter
cities. It can be considered as the most essential tool for supporting the
Internet of Things-based learning system, which is paving the way for more
efficient ways in the online-education of smart cities in the not-too-distant
future, despite its existing challenges. Future research has to stay focused on
present online learning methods and their similarity with smart cities, as
well as the creation of new approaches based on this fact.
10.2 Review of Literature

Professionals who want to further their careers and students who want to
acquire new skills while maintaining their existing work status, as well as
individuals who want to obtain training or prepare for certification exams,
are increasingly turning to online education. Institutions are finding it
increasingly challenging to keep careful monitoring and evaluation of their
courses as the number of persons enrolling in online courses grows [9].
However, owing to a shortage of supervisors and tutors, educational insti-
tutions are investing extensively in online teaching and learning to allow
their students to utilize the network as a platform. High-bandwidth appli-
cations like animations, video conferencing, and simulations, all of which
may be provided over the Internet, are available to a worldwide audience
of networked learners through the Internet [10]. The website offers a col-
laborative environment for sharing materials with others, as well as online
course courses. Sensors are utilized in the Internet of Things (IoT) to
collect data from learners when they interact with virtual platforms like
(MOOCs) and the data is then transferred to a central database system to
be analyzed. The goal of this investigation is to see whether the IoT can be
used to improve online learning and teaching. Several data-mining tech-
nologies will be deployed inside the central database systems to filter, orga-
nize, integrate, and analyze data in order to provide reports for various
management levels [11].
The word “e-learning” refers to the application of technology in the field
of education and learning. Learners may utilize virtual classrooms, which
are given via e-learning, to expand their knowledge without needing to
employ traditional learning methods. As a consequence of the fast prolif-
eration of information and communication technology in today’s environ-
ment, the letter “e” has become the symbol of the information technology
age we are currently living in [12]. The letter “e” is a frequent abbrevia-
tion for electrical components. As seen by the emergence of phrases like
“e-learning,” “e-health,” “e-business,” “e-government,” and many others, the
prefix “e” is becoming increasingly popular in every industry. As a conse-
quence of globalization, networking, and information technology reaching
their pinnacle, e-learning is becoming increasingly significant in the field
of education [13]. Today’s world is controlled by globalization, network-
ing, and IT. In eLearning, social media plays an essential role in enhancing
the learner experience. In recent years, the use of social media to promote
course material, training programs, learning tools, and registration in new
courses has grown in popularity. The most extensively used social media
sites for marketing include YouTube, LinkedIn, Twitter, Facebook, and

Google Plus. The below table gives you a detailed number of online social
platforms in India. IoT is a widely utilized technology that allows commu-
nication and collaboration through the Internet. It grows in both size and
dimension as it develops, affecting many aspects of our life, including edu-
cational chances [14]. According to the Internet of Things’ basic premise,
any devices having a unique IP address are allowed to connect with one
another in the future, both physically and digitally. The data acquired by
sensors, tags, or actuators and transferred to a cloud system through a gate-
way is the most basic structure of the IoT. In the IoT, machine-to-machine
with object-to-machine interactions are prevalent. The IoT has facilitated
the expansion of a wide range of applications, from simple domestic appli-
ances to complex medical equipment [15]. The Internet of Things covers a
broad variety of human activities, including smart cities, smart businesses,
and smart energy use to name a few.
Education is one of the most important activities of humans that the
Internet of Things has influenced, with the shape of education in the not-
too-distant future likely to be converted into a more imaginative frame-
work. IoT is a relatively new phenomenon that promotes innovation across
a broad variety of businesses. The area of education is one of these sectors
(sometimes known as e-education). IoT has the potential to deliver a broad
variety of e-educational technologies that will have a major influence on
the future of educational institutions since it can be combined with other
information technology (IT) technologies. Intelligent technologies will be
installed in the future educational facility [16]. Students and teachers must
verify their position as users by showing their finger prints and RFID ID
cards in front of the reader, as well as mobile checking, whether visiting the
rooms in person or gaining permission to the school’s automated system
administration. In the future, sensors will be deployed in IoT classrooms to
ensure that only authorized teachers and students have access to the space.
The RFID or WSN devices that will be mounted on the smart whiteboards
and desks will be able to detect and identify the persons who use them
physically. Students and teachers in the smart classroom may connect with
one another in a mutually advantageous manner. The Internet of Things
(IoT) has the potential to provide secure interaction, linking all physical
and technological objects. As a consequence of this technology, students
may connect online to labs, libraries, didactic materials, tests, and instruc-
tional messages while also doing administrative tasks more efficiently in a
large virtual classroom. Another benefit of this kind of virtual learning is
that in this format, all learning assignments and activities will be described
as objects. E-learning also refers to the incorporation of electronic tools,
software and hardware applications, and web-based activities into a learn-

ing system or system of learning, and it dates back to the 1980s. In fact, due
to the fast development of technology and communication technologies, it
is now feasible to study online using a large learning environment, which
is becoming more popular (Internet) [17].
The Internet of Things (IoT) allows learning environments to be
expanded by integrating actual and virtual things without interfering with
the learning process. Traditional e-learning has the ability to provide learn-
ers with a wide virtual access environment as a digital strategy, but it also
has important limitations, as noted below [18]. The most major constraints
in e-learning settings are geographical location, face-to-face communica-
tion between objects, and effective cooperation between virtual and real
agents. The usage of smart objects in the learning environment is one
potential answer to the issues listed above. The Internet of Things (IoT)
is commonly recognized as the most important source of smart agents for
e-learning. The Internet of Things (IoT) has the ability to include two key
aspects into traditional e-learning: intelligence and object interaction [19].
IoT has the potential to establish a vast platform for students and instruc-
tors to access a variety of distant learning devices and products. A high
degree of interaction between virtual and physical goods may lead to the
creation of a wide range of collaborative situations.
10.2.1 Objectives of the Study

The objective was to make a comparative study of an IoT based Ensemble
Predictive system with real-life teacher predictions on student observation
during E-Learning and further, to identify the forecasting efficiency of the
IoT based Ensemble Predictive system with the real world.
10.3 Methodology
The descriptive and experimental research design is adopted for further
study. The data is collected from 46 faculties for 188 periods through an
opinion-based survey using a questionnaire. Similarly, the 188 periods the
data was collected from created an IoT based Ensemble Predictive System.
The system is designed in such a way that it uses five variables, namely:
Level of Interaction, No. of Questions Raised, No. of Students in the Class,
No. of Concepts Taught in a Period, and Responsiveness of the Students to
Questions during Class Hours, to perform the student observing analysis.
10.4 Analysis and Interpretation

Analysis is done to find out the average score provided by the system and
opinions provided by the faculties during 188 hours.
Table 10.1 Group statistics - teacher & system feedback.
Group statistics
Std. Std. error
Teacher/system feedback N Mean deviation mean
Level of Teacher 188 4.3564 .69806 .05091
interaction feedback
System 188 4.4255 .70883 .05170
feedback
No. of questions Teacher 188 4.4255 .70125 .05114
raised feedback
System 188 4.3457 .70342 .05130
feedback
No. of students Teacher 188 4.3723 .66194 .04828
in the class feedback
System 188 4.4362 .60406 .04406
feedback
No. of concepts Teacher 188 4.3883 .73366 .05351
taught in a feedback
period
System 188 4.4043 .65111 .04749
feedback
Responsiveness Teacher 188 4.3989 .67452 .04919
of the feedback
students to
System 188 4.3404 .67107 .04894
questions
feedback
Source: (Primary data)
From the Table 10.1 Group statistics - teacher & system feedback, mean
score calculated using the opinion provided by the faculties and output of
the analysis made by the system, it can be interpreted that students have
a high level of observation during the class hours, as the mean score lies
between (4.3404 - 4.4255).
Analysis is done to check the difference in opinion between the faculties
and system.
From Table 10.2 Independent sample test - teacher & system feedback,
the approximated value is more than 0.05, meaning the null hypothesis is
accepted. Therefore, there is no major difference in the answers provided by
the faculties and output generated by the IoT Ensemble Predictive system.
Therefore, analysis is taken forward to identify whether there is any
significant relationship between the opinion provided by the faculties and
output generated by the IoT Ensemble Predictive system.
From Table 10.3 Correlation analysis - teacher & system feedback, the
approximate value is less than 0.05 for most of the items, meaning the null
hypothesis is rejected. Therefore, there is a closed relationship between
the opinion provided by the faculties and output generated by the IoT
Ensemble Predictive system.
But for the item (No. of Students in the Class), the evaluated value is
greater than 0.05 for most of the items, meaning the null hypothesis is
accepted. But, there is no serious relationship between the answers pro-
vided by the faculties and output generated by the IoT Ensemble Predictive
system when the No. of students increased or decreased beyond normal
ideal conditions.
Table 10.2 Independent sample test - teacher & system feedback.

Independent samples test
Levene’s
test for
equality of
variances t-test for equality of means
F Sig. t df Sig. (2-tailed)
Level of Equal .371 .543 -.953 374 .341
interaction variances
assumed
Equal variances -.953 373.912 .341

not assumed
No. of questions Equal .171 .680 1.101 374 .271

raised variances
assumed
Equal 1.101 373.996 .271

variances
not
assumed
(Continued)
Table 10.2 Independent sample test - teacher & system feedback. (Continued)
Independent samples test
Levene’s
test for
equality of
variances t-test for equality of means
F Sig. t df Sig. (2-tailed)
No. of students Equal variances .164 .686 -.977 374 .329
in the class assumed
Equal -.977 370.912 .329

variances
not
assumed
No. of concepts Equal .974 .324 -.223 374 .824

taught in a variances
period assumed
Equal -.223 368.794 .824

variances
not
assumed
Responsiveness Equal .052 .820 .843 374 .400

of the students variances
to questions assumed
Equal variances .843 373.990 .400

not assumed
Herein, an analysis was carried out to identify the forecasting efficiency

of the IoT based Ensemble Predictive system with the real world.
From Table 10.4 Regression analysis - teacher & system feedback, the
Calculated R-value is 0.788, meaning there is a 78.8% strong positive rela-
tionship between the opinion provided by the faculties and output gener-
ated by the system. The estimated R-Square Value (0.621) indicates that
it is possible to forecast an IoT based Ensemble Predictive system with in
the real world with 62.1% efficiency. Also, the ANOVA significance value
is less than 0.05, illustrating the model is fit. Furthermore, the estimated
coefficients’ significance value is less than 0.05 and this confirms that an
IoT based Ensemble Predictive system can be forecasted with real-world
information and vice-versa.
Table 10.3 Correlation analysis - teacher & system feedback.

Correlation
System feedback - level of
interaction
Teacher feedback - Pearson .384
level of correlation
interaction
Sig. (2-tailed) .000
N 188
System feedback - no. of questions
Raised
no. of questions correlation
raised
N 188
System feedback - no. of students in
the class
no. of students in correlation
the class
N 188
System feedback - no. of concepts
taught in a period
no. of concepts correlation
taught in a
period Sig. (2-tailed) .026
N 188
System feedback - responsiveness of
the students to questions
(Continued)
Table 10.3 Correlation analysis - teacher & system feedback. (Continued)

Correlation
responsiveness correlation
of the students to
questions Sig. (2-tailed) .019
N 188
Overall system feedback
Overall teacher Pearson .788
feedback correlation
N 188
10.5 Findings and Conclusion

From the evaluation, it can be interpreted that there is no big difference
in the opinion provided by the faculties and output generated by the IoT
Ensemble Predictive system. Further, it was identified that there is a sig-
nificant relationship between the opinion provided by the faculties and
output generated by the IoT Ensemble Predictive system. But for the item
(No. of Students in the Class), there is no significant relationship between
the opinion provided by the faculties and output generated by the IoT
Ensemble Predictive system. Therefore, it can be interpreted that when the
No. of students increased or decreases beyond normal ideal conditions,
there is the possibility of deviation in the opinion provided by the facul-
ties and output generated by the system. Also, The Calculated R-value is
0.788, meaning there is a 78.8% strong positive relationship between the
opinion provided by the faculties and output generated by the system. The
estimated R-Square Value (0.621) indicates that it is possible to forecast an
IoT based Ensemble Predictive system in the real world with 62.1% effi-
ciency. Furthermore, the evaluated coefficients significance value is less
than 0.05 and this confirms that IoT based Ensemble Predictive system
can be forecasted with real-world information and vice-versa.
Table 10.4 Regression analysis - teacher & system feedback.

Model summary
Std. error of
Model R R square Adjusted R square the estimate
1 .788a .621 .378 .32915
a. Predictors: (constant), overall teacher feedback
ANOVA
Sum of
Model squares df Mean square F Sig.
1 Regression 12.432 1 12.432 114.747 .000b
Residual 20.151 186 .108
Total 32.583 187
a. Dependent variable: overall system feedback
b. Predictors: (constant), overall teacher feedback
Coefficientsa
Unstandardized Standardized
coefficients coefficients
Model
B Std. error Beta t Sig.
1 (Constant) 1.961 .228 8.602 .000
Overall .554 .052 .618 10.712 .000
teacher
feedback
a. Dependent variable: overall system feedback
References
1. Berecz, “Overview of E-learning Strategies from the Point of View of Higher
Education,” J. Appl. Multimed., vol. 13, no. 4, pp. 117–127, 2019, doi:
10.26648/jam.2018.4.003.
2. Hammi, R. Khatoun, S. Zeadally, A. Fayad, and L. Khoukhi, “IoT technolo-
gies for smart cities,” IET Networks, vol. 7, no. 1, pp. 1–13, 2018, doi: 10.1049/
iet-net.2017.0163.
3. M. Njeru, M. S. Omar, S. Yi, S. Paracha, and M. Wannous, “Using IoT tech-

nology to improve online education through data mining,” Proc. 2017 IEEE
Int. Conf. Appl. Syst. Innov. Appl. Syst. Innov. Mod. Technol. ICASI 2017, no.
May, pp. 515–518, 2017, doi: 10.1109/ICASI.2017.7988469.
4. S. Kusuma and D. Kasi Viswanath, “IOT and Big data analytics in e-Learn-
ing: A technological perspective and review,” Int. J. Eng. Technol., vol. 7, no.
1, pp. 164– 167, 2018, doi: 10.14419/ijet.v7i1.8.11540.
5. M. Bayani, K. Leiton, and M. Loaiza, “Internet of things (IOT) Advantages
on e-learning in the smart cities IoT-based library automation & monitoring
system view project Internet of things (IoT) advantages on e-learning in the
smart cities,” Int. J. Dev. Res., vol. 07, no. January, pp. 17747–17753, 2017,
[Online]. Available: http://www.journalijdr.com.
6. S. Madakam, R. Ramaswamy, and S. Tripathi, “Internet of Things (IoT): A
Literature Review,” J. Comput. Commun., vol. 03, no. 05, pp. 164–173, 2015,
doi: 10.4236/jcc.2015.35021.
7. Y. S. Sai and K. K. Kumar, “Internet of things and its applications,” Int. J. Eng.
Technol., vol. 7, pp. 422–427, 2018, doi: 10.14419/ijet.v7i2.7.10758.
8. V. D. Soni, “IOT connected with e-learning,” Int. J. Integr. Educ., vol. 2, no. 5,
pp. 273–277, 2019.
9. K. Balakrishna, F. Mohammed, C. R. Ullas, C. M. Hema, and S. K. Sonakshi,
“Application of IOT and machine learning in crop protection against ani-
mal intrusion,” Glob. Transitions Proc., vol. 2, no. 2, pp. 169–174, 2021, doi:
10.1016/j.gltp.2021.08.061.
10. Abdellatif et al., “Communication-Efficient Hierarchical Federated Learning
for IoT Heterogeneous Systems with Imbalanced Data,” Futur. Gener.
Comput. Syst., vol. 128, pp. 406–419, 2021, doi: 10.1016/j.future.2021.10.016.
11. K. F. Muteba, K. Djouani, and T. O. Olwal, “Deep reinforcement learning
based resource allocation for narrowband cognitive radio-Iot systems,”
Procedia Comput. Sci., vol. 175, no. 2019, pp. 315–324, 2020, doi: 10.1016/j.
procs.2020.07.046.
12. J. Su, S. He, and Y. Wu, “Features Selection and Prediction for IoT Attacks,”
High- Confidence Comput., p. 100047, 2021, doi: 10.1016/j.hcc.2021.100047.
13. E. Anthi, L. Williams, A. Javed, and P. Burnap, “Hardening machine learn-
ing denial of service (DoS) defences against adversarial attacks in IoT smart
home networks,” Comput. Secur., vol. 108, p. 102352, 2021, doi: 10.1016/j.
cose.2021.102352.
14. M. W. Rahman, R. Islam, A. Hasan, N. I. Bithi, M. M. Hasan, and M. M.
Rahman, “Intelligent waste management system using deep learning with
IoT,” J. King Saud Univ. - Comput. Inf. Sci., no. xxxx, 2020, doi: 10.1016/j.
jksuci.2020.08.016.
15. B. Song et al., “ScienceDirect IoT and Machine learning for in-situ process
control using Laser Based IoT and Machine learning for in-situ process using
Laser Based Additive study Additive Manufacturing (LBAM) case study
new methodology and physical architecture of a , * analyze the functional
existing products for an assembly oriented Tiwari a product family iden-

tification,” Procedia CIRP, vol. 104, pp. 1813– 1818, 2017, doi: 10.1016/j.
procir.2021.11.306.
16. D. Deng, X. Li, V. Menon, M. J. Piran, H. Chen, and M. A. Jan, “Learning-
based joint UAV trajectory and power allocation optimization for secure IoT
networks,” Digit. Commun. Networks, no. August 2020, 2021, doi: 10.1016/j.
dcan.2021.07.007.
17. R. Akhter and S. A. Sofi, “Precision agriculture using IoT data analytics and
machine learning,” J. King Saud Univ. - Comput. Inf. Sci., no. xxxx, 2021, doi:
10.1016/j.jksuci.2021.05.013.
18. V. R. Kebande, R. A. Ikuesan, N. M. Karie, S. Alawadi, K. K. R. Choo, and
A. Al-Dhaqm, “Quantifying the need for supervised machine learning in
conducting live forensic analysis of emergent configurations (ECO) in IoT
environments,” Forensic Sci. Int. Reports, vol. 2, no. July, p. 100122, 2020,
doi: 10.1016/j.fsir.2020.100122.
19. M. Esmail Karar, A.-H. Abdel-Aty, F. Algarni, M. Fadzil Hassan, M. A.
Abdou, and O. Reyad, “Smart IoT-based system for detecting RPW larvae in
date palms using mixed depthwise convolutional networks,” Alexandria Eng.
J., 2021, doi: 10.1016/j.aej.2021.10.050.
11
Modelling and Analysis of a
Congestion Dependent Queue
with Bernoulli Scheduled Vacation
Interruption and Client Impatience
K. Jyothsna1* and P. Vijaya Kumar2
Department of Mathematics, Anil Neerukonda Institute of Technology

1
and Sciences, Visakhapatnam, Andhra Pradesh, India

2
Department of Mathematics, GITAM (Deemed to be University), Visakhapatnam,
Andhra Pradesh, India
Abstract
This paper’s purpose is to model and analyze an impatient client renewal input
Bernoulli scheduled vacation interruption queue. Two types of congestion depen-
dent client impatience, namely balking and reneging, have been incorporated.
The server shifts to working vacation mode whenever the system is completely
depleted. The server follows Bernoulli’s vacation interruption schedule after exe-
cuting a service during working vacation, i.e., the server either continues working
during the vacation with certain probability or interrupts the working vacation
with the complementary probability. Regular service durations, working vacation
service durations, and vacation durations are expected to be congestion dependent
and exponentially distributed. The stationary probability distributions at different
epochs were estimated utilizing supplementary variable and iterative approaches.
A few performance metrics and the model parameters’ influence on the perfor-
mance metrics have been provided through numerical outcomes.
Keywords: Balking, reneging, congestion dependent, multiple working

vacations, vacation interruption, Bernoulli schedule
*Corresponding author: mail2jyothsnak@yahoo.co.in
151
11.1 Introduction
Queues or waiting lines are prominent in normal life and even in a range of
corporate and manufacturing contexts. Waiting for service is inescapable,
resulting in dissatisfaction, annoyance, and anxiety among clients. Clients
are strained for time in today’s fast-paced environment. They would pre-
fer not to be involved in the act of waiting for service. Clients’ behavior is
affected by such time limitations. Balking and reneging are two terms used
in queuing parlance to describe this type of impatience. If a client on arrival
discovers the queueing system is not empty, he or she may elect not to join it.
Balking is the term used in queueing theory to describe this behavior. Clients
who queue up but leave before the service is completed are termed to be
reneged. Clients’ balking behavior in an exponential queueing situation has
been initially proposed by Haight [1]. For the single server Markovian prob-
lem, Haight [2] introduced the reneging behavior of clients. The research of
a balking queue under renewal input process has been conducted by Finch
[3]. Wang et al. [4] carried out an optimum and sensitivity analysis as well as
a number of numerical experiments on a machine maintenance issue includ-
ing multiple servers and balking. Ward [5] examined the outcomes as well as
the reneging behavior of the queueing problem under traditional high traffic.
Queuing models with server vacations are especially significant for laying
the groundwork for effective design and research in a number of real life
scenarios like commercial and industrial manufacturing units, software and
information networking, and other sectors. Clients cannot access the service
facility for a predefined duration of time in vacation queueing approaches.
The server may be clearing surplus tasks left by departing clients, undergoing
restoration, or taking a break during vacation. Several authors have explored
queueing scenarios where the service facility takes a break, see [6−16].
Working vacation queues have gained popularity in the modern gener-
ations as a result of possible uses in data transmission, internet technology,
industrial systems, and many other associated systems. Servi and Finn [17]
presented the idea of working vacations (WV), wherein rather than shut-
ting down altogether, the service facility delivers service to the arrivals at a
decreased rate. This process is continued until a waiting client is discovered
at a vacation termination instant. A series of such WV is stated as multi-
ple working vacations (MWV). Working vacation queues are more conve-
nient to utilize since they are more accessible than the traditional vacation
queues. Due to this, working vacation queueing systems have sparked a
lot of interest. Banik et al. [18] published the analysis of an MWV queue,
wherein the arrivals follow a renewal process, utilizing a supplementary
Modelling and Analysis of a Congestion Dependent Queue 153
variable method. A finite buffer balking queue under renewal input process
has been reported by Vijaya Laxmi and Jyothsna [19]. Majid [20] worked
on the study of client impatience in a working vacation waiting line.
When some system indices reach a specific threshold in WV queues,
the server may restore regular service, interrupting the working vacation.
This sort of action is referred to as vacation interruption (VI) in working
vacation queues. Li and Tian [21] initially came up with the idea of VI
in a Markovian queue. Li et al. [22] extended Li and Tian [21]’s research
to a VI queue wherein the arrivals follow a renewal process. Majid and
Manoharan [23] published the research of a Markovian queue with VI
wherein the server is expected to interrupt the WV if the system detects at
least N waiting clients at a service finishing moment during the WV. At the
moment the service is completed in WV, the server determines whether to
extend the WV with a probability of q or to disrupt the WV with the com-
plimentary probability of 1 − q, referred to as Bernoulli scheduled vacation
interruption. Initially, Zhang and Shi [24] explored VI under the Bernoulli
rule in an exponential queue. The investigations of an impatient client VI
queue under Bernoulli rule and renewal input process has been published
by Vijaya Laxmi et al. [25]. More information regarding working vacation
queues with client impatience may be found in Bouchentouf and Yahiaoui
[26], Majid et al. [27], as well as references therein.
The majority of queueing system research is predicated on the notion
that the system’s state has no effect on the input and service parameters. In
practice, however, many of the queueing system’s service and input param-
eters are based on the system’s state and reducing the clients wait time.
Consider the case of telecommunications network routing. Congestion
independent routing probabilities cannot prevent tasks from being sent
to already overburdened disc stations in such networks, whereas jobs are
routed to the least loaded discs first when congestion dependent routing
probabilities are used. State dependent routing outperforms state inde-
pendent routing in terms of throughput. For literature on state dependent
queues, see [28−30] and the references therein.
To our knowledge, no research has ever examined queues that include
congestion dependence, working vacations, Bernoulli schedule VI, and
impatient clients; this gap in the literature drives the current research.
Also, the uncorrelated arrival process often yields a better approxima-
tion than the exponential distribution since the memoryless quality of the
aforementioned distribution might not necessarily fit the aims of applica-
tions and it can incorporate Erlang, deterministic distributions, etc. Hence,
investigating the impact of client impatience in a congestion dependent
Bernoulli scheduled VI queue under renewal input process is worthwhile.
Service, as well as the vacation durations of the service facility, are expected
to be exponentially distributed and congestion dependent. The concept is
researched employing supplementary variable and iterative techniques.
The stationary distribution of the number of clients has been computed at
pre-arrival and arbitrary epochs. Performance metrics and numerical data
indicating the change in the performance metrics with the model parame-
ters are provided in table and graph forms.
The following is the layout of the article’s main body. Section 11.2 pres-
ents the model’s overview. The computation of the stationary distributions
at diverse time epochs is in Section 11.3 and special cases of the model
in Section 11.4. In Section 11.5, several model performance metrics are
provided and numerical calculations to show the parametric influence on
system performance metrics are disposed in Section 11.6. Finally, Section
11.7 brings the article to a conclusion.
11.2 Model Overview

A GI/M/1/N/MWV queue that is congestion dependent and having balking,
reneging, vacation interruption under a Bernoulli schedule is studied. The
time between two subsequent arrivals are assumed to be distributed identi-
cally and are independent random variables with probability density function
a(z), z ≥ 0, Laplace-Stieltjes transformation (LST) Â(θ ), and mean inter-ar-
rival duration 1/λ = −Aˆ (1)(0). When a client views j clients in the system, it
can queue up with probability βj or balk with the complimentary probability
1 − βj. We also presume that β0 = 1, 0 ≤ βj+1 < βj ≤ 1, 1 ≤ j ≤ N – 1, and βN = 0.
If an arrival does not join the queue, he or she will be permanently lost.
If an arrival has to wait longer than expected for service after entering
the queue, he may renege. With parameter α, the client’s waiting period pre-
ceding reneging is exponentially distributed. (j − 1)α is taken as the average
reneging rate, as an impatient client’s arrival and departure are unrelated.
A single service facility offers services on a first-come, first-served
(FCFS) basis. The service facility departs for a WV as and when the system
is depleted. Unlike the queueing paradigms with conventional vacation
policy, the service is provided to the waiting clients at a slower rate during
WV. At the moment, the service is completed during WV, the server deter-
mines whether to extend the vacation with the probability of q or to disrupt
the WV with a complimentary probability of 1 − q, designated as Bernoulli
scheduled vacation interruption.
When there are j clients in the system, the regular service durations,
WV service durations, and vacation durations are all considered to be
exponentially distributed with congestion dependent rates µj, ηj and ϕj, 1 ≤

j ≤ N, respectively. Further, define the average regular service rate (µm),
average vacation service rate (ηm), and average vacation rate (ϕm) as
∑ ∑ ∑
N N N
µj ηj φj
=j 1 =j 1 j =1
=µm = ; ηm ; φm =
N N N
At time t, let
• Ns(t) = The total count of clients, counting the one who is

currently undergoing service
• Z(t) = Upcoming arrival’s remaining inter-arrival time.
At steady state, define:
π j (z )dz lim
= = Pr {N s (t ) j, z ≤ Z (t ) < z + dz}, z ≥ 0, 0 ≤ j ≤ N ,
t →∞
ω j (z )dz = lim Pr {N s (=
t ) j, z ≤ Z (t ) < z + dz }, z ≥ 0, 1 ≤ j ≤ N ,
t →∞
The Laplace-Stieltjes transformations that follow are introduced:

∞
=π j (θ )
∫ e −θ zπ j (z )dz , 0 ≤ j ≤ N ,
^
0
∞
=ω j (θ )
∫ e −θ zω j (z )dz 1 ≤ j ≤ N .
^
At an arbitrary instant, while the server is in WV and regular service, the

probability of j clients in the system can be denoted by π j ≡ πˆ j (0),(0 ≤ j ≤ N )
and ω j ≡ ωˆ j (0),(1 ≤ j ≤ N ), respectively.
11.3 Model Analysis

To determine the system length probabilities at arbitrary instants, differ-
ential difference equations are formulated that associate the distribution
of number of clients in the system at the completion of the vacation and
regular service period. We accomplish this by employing the supplemen-
tary variable technique, which connects the system’s state at two time
instants t and t+dt. The differential difference equations under steady-state
conditions can be constructed utilizing probabilistic reasonings and taking

limit t → ∞ as
−π 0(1) (z ) =
µ1ω1(z ) + η1π 1(z ),
−π (j1) (z ) =
−r ( j)π j (z ) + a(z )((1 − β j )π j (0) + β j−1π j−1(0)) + (qη j+1
+ jα )π j+1(z ), 1 ≤ j ≤ N − 1,
−π N(1) (z ) =
−r (N )π N (z ) + a(z )(π N (0) + β N −1π N −1(0)),
−ω1(1) (z ) =
−s(1)ω1(z ) + a(z )(1 − β1 )ω1(0) + φ1π 1(z ) + s(2)ω2 (z )qη2π 2 (z ),
−ω (j1) (z ) =
−s( j)ω j (z ) + a(z )((1 − β j )ω j (0) + β j−1ω j−1(0)) + φ jπ j (z )
+ qηπ j+1(z ) + s( j + 1)ω j+1(z ), 2 ≤ j ≤ N − 1,
−ωN(1) (z ) =
−s(N )ωN (z ) + φN π N (z ) + a(z )(β N −1ωN −1(0) + ωN (0)),
where πj(0) and ωj(0) are respective rates of arrivals and for 1 ≤ j ≤ N, r(j) =
ϕj + ηj +(j − 1)α, s(j) = µj + (j − 1)α. The preceding set of equations is mul-
tiplied with e−θz and integrated over z for 0 to ∞ to get
−θ π 0 (θ ) =
−π 0 (0) + µ1 ω1(θ ) + η1 π 1(θ ),
^ ^ ^
(11.1)
^
−π j (0) + A(θ )((1 − β j )π j (0) + β j −1π j −1(0))
(r ( j) − θ )π j (θ ) =
^
+ (qη j +1 + jα )π j + 1(θ ),
^
1 ≤ j ≤ N – 1, (11.2)
(r ( N ) − θ )πˆ N (θ ) = −π N (0) + Aˆ (θ )(π N (0) + β N −1π N −1 (0)), (11.3)
^
(s(1) − θ ))ω1(θ ) =
−ω1(0) + A(θ )(1 − β1 )ω1(0) + φ1 π 1(θ )
^ ^
+ s(2)ω 2 (θ ) + qη2π 2 (θ ),
^
(11.4)
^
(s( j) − θ )ω j (θ ) =
−ω j (0) + A(θ )((1 − β j )ω j (0) + β j−1ω j−1(0))
^
+ φ j π j (θ ) + qηπ j+1(θ )
^
+ s( j + 1)ωˆ j +1 (θ ),2 ≤ j ≤ N − 1, (11.5)

(s( N ) − θ )ωˆ N (θ ) = −ω N (0) + φ N πˆ N (θ ) + Aˆ (θ )( β N −1ω N −1 (0) + ω N (0)).

(11.6)
Adding the above equations results as:
1 − Aˆ (θ )
∑ ∑ ∑ ∑
N N N N
πˆ j (θ ) + ωˆ j (θ ) = π j (0) + ω j (0 )
j =0 j =1 θ j =0 j =1
∑
N
Assigning θ → 0 and utilizing the normality condition πj +
j =0
∑
N
ω j = 1 produces
j =1
∑ ∑
N N
π j (0) + ωj =
λ (11.7)
=j 0=j 1
11.3.1 Pre-Arrival Epoch Probabilities

Let π j− , 0 ≤ j ≤ N , and ω j− ≤ j ≤ N symbolize the probability of j cli-
ents in the system during WV and during regular service period, respec-
tively, at pre-arrival instant. Applying Bayes’ theorem, we have
1 1
=π j− = π j (0); ω j− ω j (0). (11.8)
λ λ
To obtain the pre-arrival instant probabilities (πˆj, ωˆj), we need to evalu-

ate the rate probabilities (πj(0), ωj(0)) as shown below.
Taking θ = r(N) in Equation (11.3) gives
πN−1(0) = ϱN−1πN(0), (11.9)
^
1 − A(r (N ))
where ϱN = 0, ϱN−1 = ^ . Substituting πN−1 (0) in (11.3), we get
A(r (N ))
π N (θ ) = Ψ N ,θ π N (0),
^
where
^
{A(θ )(β N −1N −1+N )−N
Ψ N ,θ = , if θ ≠ r (N ),
r (N ) − θ
^ (1)
− A (θ )(β N −1N −1 + N ), if θ =
r (N )
with
^ (l )
{A (θ )(β N −1N −1+N ) + lψ l −1
Ψ N ,θ = , if θ ≠ r (N ),
r (N ) − θ
^ (1)
− A (θ )(β N −1N −1 + N ), if θ =
r (N )
Taking θ = r(j) in (11.2) and proceeding as above, we obtain
πj(0) = ϱjπN (0), j = N − 2, N − 3, . . . , 0, (11.10)
where
 j +1 (1 − Aˆ (r ( j + 1))(1 − β j +1 )) − (qη j +2 + ( j + 1)α )Ψ j +2,r ( j + 1)

j =
A(r ( j + 1))β j
and
πˆ j (θ ) = Ψ j ,θ π N (0), j = N − 1, N − 2,…,1,
where
^
A(θ )(β j−1 j−1 + (1 − β j ) j ) + (qη j+1 + jα )Ψ j+1,θ −j
Ψ j ,θ = , if θ ≠ r ( j)
r (N ) − θ
^ (1)
− ( A (θ )(β j−1 j−1 + (1 − β j ) j ) + (qη j+1 + jα )Ψ (j1+)1,θ ), if θ = r ( j)
with
A(θ )(β j −1 j −1 + (1 − β j ) j ) + (qη j +1 + jα )Ψ (jl+)1,θ + lΨ (jl,θ−1)

Ψ j ,θ = , if θ ≠ r ( j)
r ( j) − θ
∧ ( l +1)
A (θ )(β j −1 j −1 + (1 − β j ) j ) + (qη j +1 + jα )Ψ ((lj++11),θ )
− , if θ = r ( j)
l +1
Taking θ = s(N) in (11.6), we have ωN−1(0) in terms of πN (0) and ωN (0)

as
ωN−1(0) = τN−1(0)ωN(0) + νN−1πN(0), (11.11)
where
^
1 − A(s(N )) φ Ψ
τ N 1,=
= ν N 0,τ N −1 = ^ , ν N −1 = − ^ N N ,s ( N )
A(s(N ))β N −1 A(s(N ))β N −1
Substituting (11.11) in (11.6) gives
ω N (θ ) ∆ N, θ ωN (0) + Ω N ,θ π N (0),
^
=
where
{A(θ )(β N −1τ N −1 + τ N ) − τ N

∆ N ,θ = if θ ≠ s(N )
s( N ) − θ
∧ (1)
− A (θ )(β N −1τ N −1 + τ N ), if θ =
s(N ),
∧
A(θ )(β N −1ν N −1 + φN Ψ N ,θ )

Ω N ,θ = if θ ≠ s(N ),
s( N ) − θ
∧
(1)
− ( A (θ )β N −1ν N −1 + φN Ψ (N1),θ ), if θ = s(N ),
with
∧
(l )
(l ) A (θ )(β N −1τ N −1 + τ N ) − l∆(Nl −,θ1)
∆ N ,θ = if θ ≠ s(N ),
s( N ) − θ
∧ (l +1)
A (θ )(β N −1τ N −1 + τ N )
− , if θ = s(N ),
l +1
∧
(l )
(l ) A (θ )(β N −1ν N −1 + φN Ψ (Nl ),θ + lΩ(Nl −,θ1) )
Ω N ,θ = if θ ≠ s(N ),
s( N ) − θ
∧ (l +1)
A (θ )(β N −1ν N −1 + φN Ψ (Nl +,θ1) )
− , if θ = s(N )
l +1
From Equation (11.5), we get ωj(0), ωˆ j (θ ) in terms of πN (0) and ωN (0) as
ωj−1(0) = τj−1ωN(0) + νj−1πN(0), j = N − 1, . . . , 2, (11.12)
where for 2 ≤ j ≤ N − 1,
(1 − Aˆ (s( j ))(1 − β j ))τ j − s( j + 1) j +1, s ( j )

τ j −1 =
Aˆ (s( j ))β j −1
ˆ ( s( j ))(1 − β j ))ν j − s( j + 1)Ω j +1,s ( j ) − φ j Ψ j ,s ( j ) − q − η j +1Ψ j +1,s ( j )
(1 − A
ν j −1 =
Aˆ (s( j ))β j −1
and
ω̂
ω j (θ ) ∆ j, θ ω N (0) + Ω j ,θ π N (0), =
= j N − 1,…, 2,
where for 2 ≤ j ≤ N − 1,
∧ ∧
s( j + 1)∆ j +1,θ + A(θ )(β j −1τ j −1 − (1 − A(θ )(1 − β j ))τ j

∆ j ,θ = , if θ ≠ s( j),
s( j ) − θ
∧ ∧
(1) (1)
∆ j ,θ = −s( j + 1)∆(j1+)1,θ + A (θ )(β j −1τ j −1 + A (θ )(1− β j ))τ j , if θ =
s( j),
∧ ∧
s ( j + 1)Ω j + 1 ,θ + A(θ ) β j −1ν j −1 − (1 − A(θ )(1 − β j ))ν j + φ j Ψ j ,θ + q η j + 1 Ψ j + 1 ,θ

Ω j ,θ = , if θ ≠ s ( j ),
s( j) − θ
(1) ∧ (1) ∧ (1) (1) (1)

− s ( j + 1)Ω j + 1 ,θ + A (θ )( β j −1ν j −1 + A (θ )(1 − β j ))ν j + φ j Ψ j ,θ + q η j + 1 Ψ j + 1 ,θ , if θ = s ( j ),
with
∧
(l )
s( j + 1)∆(jl+)1,θ + A (θ )(β j −1τ j −1 + (1 − β j )τ j ) + l ∆(jl,θ−1)
∆ j ,θ = , if θ ≠ s( j),
s(l) − θ
∧
(l +1)
s( j + 1)∆(jl++11,θ) + A (θ )(β j −1τ j −1 + (1 − β j )τ j )
− , if θ = s( j),
l +1
(l ) ∧ (l ) (l ) (l )
(l ) s( j + 1)Ω j +1,θ + A (θ )( β j −1ν j −1 + (1 − β j )ν j ) + φ j Ψ j ,θ l Ω l j ,θ + q η j +1 Ψ j +1,θ
Ω j ,θ = , if θ ≠ s( j ),
s( j ) − θ
( l +1 ) ∧ ( l +1 ) ( l +1 ) ( l +1 )
s( j + 1)Ω j +1,θ + A (θ )( β j −1ν j −1 + (1 − β j )ν j ) + φΨ j ,θ + q η j +1 Ψ j +1,θ
− , if θ = s( j ),
l +1
Setting θ = s(1) in (11.4) yields ωN (0) in terms of πN (0) as
ωN(0) = kπN(0),
where
ν ( A(s(1))(1 − β1 ) − 1) + s(2)Ω2, s(1) + φ1Ψ1, s(1) + qη2 Ψ 2, s(1)

k= 1
τ 1(1 − A(s(1)(1 − β1 )) − s(2)∆ 2, s(1)
Lastly, (11.7) yields the remaining unknown πN(0) as
∑ ∑
N N
π N (0) λ (
= j + (kτ j +ν j ))−1
j =0 j =1
Theorem 11.1. The probabilities π −j (0 ≤ j ≤ N ) and ω −j (1 ≤ j ≤ N ) at pre-

arrival epoch are determined as follows:
∑ ∑ (kτ +ν ))
N N
=π j−  j ( j + j j
−1
,0 ≤ j ≤ N,
j =0 j =1
+ν )(∑  + ∑ (kτ
N N
−
ω
= j (kτ j j j j +ν j ))−1 ,1 ≤ j ≤ N .
j =0 j =1
Proof. Using (11.8) and (11.13) in (11.9) to (11.12), we obtain the result
of the theorem.
11.3.2 Arbitrary Epoch Probabilities

In this subsection, the pre-arrival probabilities π −j and ω −j are used to
derive the arbitrary epoch probabilities πj (0 ≤ j ≤ N ) and ωj (1 ≤ j ≤ N ).
This is treated as a theorem given below.
Theorem 11.2. The probabilities at the arbitrary epoch are obtained as
(λβ N −1 ) −
πN = π N −1 ,
r (N )
1
=πj ((qη j+1 + jα )π j+1 + λ (β j−1π j−−1 − β jπ j− )), =j N − 1,. . .,1,
r ( j)
1
=ωN (φN π N + λβ N −1ωN− −1 ),
s( N )
1
=ωj (s( j + 1)ω j+1 + φ jπ j + qη jπ j+1 + λ (β j−1π j−−1 − β jπ j− )),
s( j )
=j N − 1,. . ., 2
1
ω1 = (s(2)ω2 + φ1π 1 + qη2π 2 − λβ1π 1− )
s(1)
N
π0 = 1− ∑(π
j =1
j + ω j ).
Proof. The intended outcome of the theorem is obtained by placing

θ = 0 in Equations (11.2) to (11.6), utilizing (11.8) and the normalization
condition.
11.4 Special Cases

Case 1: Taking q = 1, α = 0, and βj = 1 for 1 ≤ j ≤ N − 1, the present model
reduces to GI/M (n)/1/N queue with MWV and the findings are consistent
with Goswami et al. [28].
Case 2: Setting q = 1, µj = µ, ηj = η, ϕj = ϕ ∀ j = 1, 2, . . . , N and setting α =
0, the current model is reduced to a GI/M/1/N/MWV queue with balk-
ing. Our findings are identical to Vijaya Laxmi and Jyothsna [19] in this
situation.
Case 3: Taking q = 1, the current model gets reduced to GI/M (n)/1/N/
MWV queue with balking and reneging and the outcomes are consistent
with Vijaya Laxmi and Jyothsna [29].
Case 4: Taking µj = µ, ηj = η, ϕj = ϕ ∀ j = 1, 2, . . . , N brings down our model
to a GI/M/1/N/MWV queue with balking, reneging, and Bernoulli sched-
uled VI. The results are consistent with Vijaya Laxmi et al. [25].
Case 5: Taking q = 1, µj = µ, ηj = η, ϕj = ϕ, βj = 1 ∀ j = 1, 2, . . . , N and α =
0 lowers the current model to GI/M/1/N/MWV queue. The results match
with Banik et al. [18].
Case 6: Setting q = 1, µj = µ, ηj = 0, ϕj = ϕ ∀ j = 1, 2, . . . , N and taking α = 0,
βj = 1 for 1 ≤ j ≤ N, our numerical findings correspond to GI/M/1/N queue
with multiple vacations (Karaes-man and Gupta) [7].
Case 7: Taking q = 1, µj = µ, ηj = 0, ϕj → ∞ ∀ j = 1, 2, . . . , N and setting α =
0 the current model becomes a renewal input balking queue. The working
vacation probabilities πj(1 ≤ j ≤ N) do not occur in this situation and our
outcomes agree with Finch [3] (GI/M/1/N queue with balking).
11.5 Performance Metrics

The performance metrics of the model can be assessed upon obtaining
the state probabilities at pre- arrival and arbitrary instants. The expected
system length (Ls), the blocking probability (Pl), and the expected waiting
duration in the system (Ws) of a client employing Little’s Rule are, respec-
tively, given by:
Ls = ∑ j(π + ω ); P = π
j =1
j j l
−
j + ω −j ; Ws = Ls /λe ,
where λe = 1 − Pl represents the effective arrival rate.

The mean rate of balking (br), the mean rate of reneging (rr), and the
mean rate of client loss (lr) are measured by
N N
br = ∑ λ(1 − β )(π + ω ); rr = ∑ ( j − 1)α (π + ω ); lr = br + rr.

j =1
j j j
j =1
j j
11.6 Numerical Outcomes

This section shows how the performance metrics respond to the changes in
model parameters. The model parameters are selected at random as N = 10,
λ = 1.5, α = 0.7, and q = 0.4 and the balking function is considered as βj =
1−j/N 2 with the assumption that β0 = 1 and βN = 0. The congestion depen-
dent rates are assumed to be µj = 0.4j, ηj = 0.2j, and ϕj = 0.1j with averages
µm = 2.2, ηm = 1.1, and ϕm = 0.55, respectively. For HE2 inter-arrival time
distribution, we have chosen σ1 = 0.149883, σ2 = 0.850117, λ1 = 0.484237,
and λ2 = 2.380330.
In Table 11.1, we have presented the values of Ls, Ws, Pl, br, rr, and lr for
the following balking functions: βj = 1 − j/N2, βj = 1/(j + 1), and βj = e−j. The
mean balking rate (br) is lowest for the balking function βj = 1 − j/N2, as
Table 11.1 Numerical values of performance metrics for different

balking functions.
βj = 1 − j/N2 βj = 1/(j + 1) βj = e−j
Ls 2.008288 1.343753 1.191490
Ws 1.338865 0.895836 0.794326
Pl 0.000004 0.000000 0.000000
br 0.030131 0.772889 0.940714
rr 0.754335 0.314381 0.214521
lr 0.784466 1.087270 1.155234
noticed in the table, which in turn results in least mean rate of client loss
(lr). This supports our decision to use βj = 1 − j/N 2 as the balking function.
Figure 11.1 and Figure 11.2 show the influence of λ on the expected
system length (Ls) and the mean rate of client loss (lr) for exponential inter-
arrival time distribution, respectively, for different q. The images show that
as λ is increased, both Ls and lr grow for any q. Furthermore, for a fixed λ,
2.6
2.4 q=0.0
q=0.4
2.2 q=0.8
q=1.0
2
1.8
Ls
1.6
1.4
1.2
0.8
0.5 1 1.5 2
λ
Figure 11.1 Influence of λ on Ls.
1.8
q=0.0
q=0.4
1.6
q=0.8
q=1.0
1.4
1.2
lr
0.8
0.6
0.4
1 1.5 2 2.5
λ
Figure 11.2 Effect of λ on lr.

Ls and lr rise as q increases, meaning that Ls and lr are the least in VI queues
when compared to queues with no vacation interruption.
The impact of mean regular service rate (µm) on the expected system
length (Ls) and on the expected waiting duration (Ws) are displayed in
Figure 11.3 and Figure 11.4, respectively, for different mean vacation rates
(ϕm), wherein the inter-arrival durations are considered to be Erlang-5 dis-
tributed. As expected, Ls and Ws deplete with the increase of µm for any ϕm.
A similar effect is seen on Ls and Ws with the increase of ϕm for fixed µm.
Figures 11.5 and 11.6 show the changes in lr and Ws, respectively, while
the inter-arrival times are distributed according to deterministic distribu-
tion for different ηm and ϕm. It is worth noting that as ηm and ϕm grow, lr
and Ws deplete.
For different inter-arrival time distributions, Figure 11.7 displays the
system capacity (N) versus the blocking probability (Pl). The blocking
probability (Pl) is lowest when the inter-arrival durations are distributed
according to deterministic distribution and highest when they are distrib-
uted according to HE2 distribution.
When inter-arrival durations are exponentially distributed, the impact
of the reneging rate (α) on the expected system length (Ls) and on the mean
rate of client loss (lr) is depicted in Figure 11.8. As visualized in the graph,
the expected system length falls as α grows, but the mean rate of client loss
increases with α, as would be predicted.
2.25
φm = 0.55
2.2 φm = 0.82
φm = 1.10
2.15
2.1
2.05
Ls
1.95
1.9
1.85
1.7 1.8 1.9 2 2.1 2.2 2.3 2.4 2.5 2.6 2.7
µm
Figure 11.3 Impact of µm on Ls.

1.46
φm=0.55
1.44 φm=0.82
φm=1.10
1.42
1.4
1.38
1.36
Ws
1.34
1.32
1.3
1.28
1.26
1.7 1.8 1.9 2 2.1 2.2 2.3 2.4 2.5 2.6 2.7
µm
Figure 11.4 Impact of µm on Ws.
0.81
φm = 0.55
0.8
φm = 0.82
φm = 1.10
0.79
0.78
0.77
lr
0.76
0.75
0.74
0.73
0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6
ηm
Figure 11.5 Change in lr with ηm.

1.4
φm = 0.55
1.39
φm = 0.82
φm = 1.10
1.38
1.37
1.36
Ws
1.35
1.34
1.33
1.32
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6
ηm
Figure 11.6 Change in Ws with ηm.
0.12
HE2
Exponential
0.1
Erlang 3
Deterministic
0.08
Pl
0.06
0.04
0.02
0
4 5 6 7 8 9 10 11 12 13 14 15
N
Figure 11.7 N versus Pl.

3.5
Ls
3 Ir
2.5
1.5
0.5
0
0.5 1 1.5 2 2.5
α
Figure 11.8 α versus Ls and lr.
11.7 Conclusion
The current research work examines a finite capacity renewal input con-
gestion dependent queue with Bernoulli scheduled vacation interruption
and client impatience. The regular service rates, working vacation service
rates, and vacation rates are considered to be exponentially distributed and
are congestion dependent. To obtain the stationary probabilities at various
epochs, we have adopted supplementary variable and iterative techniques.
We also presented a few numerical simulations to show the consequence of
the model parameters on the key metrics, which are provided in table and
graph forms. The current findings would be beneficial and significant for
modelling the transportation business, manufacturing systems, medical
administration, etc. The future direction of this work would be to expand
the model under examination by incorporating state dependent inter-
arrival times.
References
1. F. A. Haight, Queueing with balking (1957), Biometrika, vol.44, issue.3-4,
pp.360-369, https://doi.org/10.1093/biomet/44.3-4.360
2. F. A. Haight, Queueing with reneging (1959), Metrika, vol.2, isuue.1, pp.186-

197, https://doi.org/10.1007/BF02613734
3. Finch, P.D., Balking in the queueing system GI/M/1 (1959), Acta Mathematica
Hungarica, vol.10, issue.1, pp.241-247, https://doi.org/10.1007/BF02063302
4. Kuo-Hsiung Wang, Yuh-Ching Liou and Dong-Yuh Yang, Cost optimization
and sensitivity analysis of the machine repair problem with variable serv-
ers and balking (2011), Procedia–Social and Behavioral Sciences, vol.25,
issue.14, pp.178-188, DOI: 10.1016/j.sbspro.2011.10.539
5. Amy R. Ward, Asymptotic analysis of queueing systems with reneging: A
survey of results for FIFO, single class models, (2012) Surveys in Operations
Research and Management Science, vol. 17, issue.1, pp.1-14, https://doi.
org/10.1016/j.sorms.2011.08.002
6. Takagi, H. (1991). Queueing Analysis: A Foundation of Performance
Evaluation, Vacation and Priority Systems, Vol. 1. Elsevier, Amsterdam.
7. Fikri Karaesmen, Surendra M. Gupta . The finite capacity GI/M/1 queue with
server vacations (1996), Journal of the Operational Research Society, vol.47,
issue.6, pp.817-828, https://doi.org/10.1057/jors.1996.101
8. Naishuo Tian, Zhe George Zhang (2006). Vacation Queueing Models:
Theory and Applications. International Series in Operations Research and
Management Science, Springer, New York.
9. Eitan Altman, Uri Yechiali, Analysis of customers’ impatience in queues
with server vacations (2006), Queueing Systems, vol.52, issue.4, pp.261-279,
https://doi.org/10.1007/s11134-006-6134- x
10. Dequan Yue, Wuyi Yue, Yanping Sun, ‘Performance analysis of an M/M/c/N
queueing system with balking, reneging and synchronous vacations of par-
tial servers’ in 2006 The Sixth Inter- national Symposium on Operations
Research and Its Applications, Xinjiang, China, ORSC and APORC, 128-143.
11. Uri Yechiali, Queues with system disasters and impatient customers when
system is down (2007), Queueing Systems, vol.56, pp.195-202, http://dx.doi.
org/10.1007/s11134-007-9031-z
12. Eitan Altman, Uri Yechiali, Infinite-server queues with system’s additional
task and impatient customers (2008), Probability in the Engineering and
Information Sciences, vol.22, issue.4, pp.477-493, http://dx.doi.org/10.1017/
S0269964808000296
13. Nir Perel, Uri Yechiali, Queues with slow servers and impatient customers
(2010), European Journal of Operational Research, vol.201, issue.1, pp.247-
258, doi:10.1016/j.ejor.2009.02.024
14. Dequan Yue, Wuyi Yue, Zsolt Saffer and Xiaohong Chen, Analysis of an
M/M/1 queueing system with impatient customers and a variant of multiple
vacation policy (2014), Journal of Industrial and Management Optimization,
vol.10, issue.1, pp.89-112, http://dx.doi.org/10.3934/jimo.2014.10.89
15. Amina Angelika Bouchentouf, Abdelhak Guendouzi, Cost optimization
analysis for an MX/M/c vacation queueing system with waiting servers and
impatient customers (2019), SeMA Journal, vol.76, issue.2, pp.309-341,

http://dx.doi.org/10.1007
16. M. I. G. Suranga Sampath, Jicheng Liu, Impact of customers’ impatience on
an M/M/1 queueing system subject to differentiated vacations with a waiting
server (2020), Quality Technology and Quantitative Management, vol.17,
issue.2, pp.125-148, https://doi.org/10.1080/16843703.2018.1555877
17. L. D. Servi, S. G. Finn, M/M/1 queue with working vacations (M/M/1/
WV) (2002), Performance Evaluation, vol.50, issue.1, pp.41–52, https://doi.
org/10.1016/S0166-5316(02)00057-3
18. A. D. Banik, U. C. Gupta, S. S. Pathak, On the GI/M/1/N queue with
multiple working vacations - analytic analysis and computation (2007),
Applied Mathematical Modelling, vol.31, issue.9, pp.1701-1710, https://doi.
org/10.1016/j.apm.2006.05.010
19. P. Vijaya Laxmi, K. Jyothsna, Analysis of finite buffer renewal input queue
with balking and multiple working vacations (2013), OPSEARCH, vol.50,
issue.(4), pp. 548-565. http://dx.doi.org/10.1007/s12597-013-0123-8
20. Shakir Majid, Performance analysis of a Markovian queue with impatient
customers and working vacation (2021), Journal of the Operations Research
Society of China, https://doi.org/10.1007/s40305-021-00361-w.
21. Jihong Li, Naishuo Tian, The M/M/1 queue with working vacations and
vacation interruption (2007), Journal of Systems Science and Systems
Engineering, vol.16, pp.121-127, https://doi.org/10.1007/s11518-006-5030-6
22. Ji-HongLi, Nai-Shuo Tian, Zhan-You Ma, Performance analysis of GI/M/1
queue with working vacations and vacation interruption (2008), Applied
Mathematical Modelling, vol.32, issue.12, pp.2715-2730, https://doi.
org/10.1016/j.apm.2007.09.017
23. Shakir Majid and P. Manoharan, Analysis of an M/M/1 queue with work-
ing vacation and vacation interruption (2019), Applications and Applied
Mathematics: An International Journal, vol.14, issue.1, pp.19-33,
24. Hongbo Zhang, Dinghua Shi, The M/M/1 queue with Bernoulli-schedule-
controlled vacation and vacation interruption (2009), International Journal
of Information and Management Sciences, vol.20, issue.(4), pp.579-587.
25. Pikkala Vijaya Laxmi, Singuluri Indira and Kanithi Jyothsna, Ant colony
optimization for optimum service times in a Bernoulli schedule vacation
interruption queue with balking and reneging (2016), Journal of Industrial
and Management Optimization, vol.12, issue.4, pp.1199–1214, http://dx.doi.
org/10.3934/jimo.2016.12.1199
26. Bochentouf, A. A., Yahiaoui, L. (2017). On feedback queueing system with
reneging and retention of reneged customers, multiple working vacations and
Bernoulli schedule vacation interruption, Arabian Journal of Mathematics,
6(1), 1-11.
27. Shakir Majid, P. Manoharan and A. Ashok, Analysis of an M/M/1 queue-
ing system with working vacation and impatient customers, American
International Journal of Research in Science, Technology, Engineering &

Mathematics (2019), Special Issue, pp.314-322,
28. V. Goswami, P. Vijaya Laxmi, K. Jyothsna, Analysis of GI/M (n)/1/N queue
with state-dependent multiple working vacations (2013), OPSEARCH,
vol.50, issue.1, pp.106-124, http://dx.doi.org/10.1007/s12597-012-0096-z
29. Pikkala Vijaya Laxmi and Kanithi Jyothsna, On renewal input state depen-
dent working vacations queue with impatient customers (2015), International
Journal of Mathematics in Operational Research, vol.7, issue.6, pp.661–680.
30. Doo Il Choi, Dae-Eun Lim, Analysis of the state-dependent queueing model and
its application to battery swapping and charging stations (2020), Sustainability,
vol.12(6), issue.2343, pp.1-15, https://doi.org/10.3390/su12062343
12
Resource Allocation Determines Alternate
Cell Fate in Bistable Genetic Switch
Priya Chakraborty and Sayantari Ghosh*
Department of Physics, National Institute of Technology Durgapur, Durgapur,

West Bengal, India
Abstract
Living cells need a constant availability of certain resources to have a sustained gene
expression process. Limited availability of cellular resources for gene expression,
like ribosomes and RNA Polymerase, significantly modifies the system dynamics.
Factors like the variation in rate of binding or variation in efficiency of the recruited
resource have the potential to affect crucial dynamic phenomena like cell fate deter-
mination. In this paper we have taken a very important motif, a bistable genetic tog-
gle switch, and explored the effect of resource imbalance in this circuit in terms of
the bifurcations taking place. We show that initial asymmetric biasing to resources
via resource affinity or gene copy number significantly modifies the cell fate tran-
sition, both in pitchfork and saddle node type bifurcation. Our study establishes
that in a limited resource environment, controlled resource allocation can be an
important factor for robust functioning of the synthetic or cellular genetic switches.
Keywords: Gene regulation, resource allocation, cell fate decision,

pitchfork bifurcation, saddle node bifurcation, genetic toggle switch
12.1 Introduction
Proteins that govern all functionalities in a living cell are produced by two
major steps: transcription and translation, which indeed are a combination
of several intermediate processes. In order to understand intriguing cellu-
lar processes like cellular decision making or to operate a synthetic circuit
inside the cell, a detailed mathematical study of cellular or synthetic gene
*Corresponding author: sayantari.ghosh@phy.nitdgp.ac.in
173
regulatory dynamics is of prime concern. Extreme non-linearity inside

cellular systems and the nature of coupling of one ongoing process with
another makes this understanding of the dynamics extremely difficult.
However, some re-occurring genetic pattern, called a motif, can be found
in living organisms performing some typical tasks for the cell. Scientists
are focusing largely to explore repeatedly occurring motifs to understand
the cell dynamics as well as cellular decision making from the last few
decades. Genetic toggle is one of the most extensively studied biological
motifs [7, 16, 18], where two genes (lacI and tetR) mutually repress each
other’s expression. Dynamics of this genetic motif are usually studied from
the perspective of saddle-node bifurcation. Here, in the case of saddle node
bifurcation of toggle, the system undergoes a transition to a bistable state
from a monostable one and after a specific range of parameters, again, the
system becomes monostable. Thus, a specific region of bistability separates
two otherwise monostable regions in the phase plane. The system can also
undergo a pitchfork bifurcation for some specific symmetry, as studied in
some recent works [2]. In case of pitchfork type bifurcation, a monostable
region converts to a bistable one at a transition point and remains bistable
for rest. Bistabilty introduces some irreversibility in biological systems,
that once the system attains its steady state, it retains its same steady state
even on application of some output perturbations on a small scale. Thus, in
the case of cell fate differentiation and cellular decision making, the genetic
toggle motif is taken as a canonical circuit for understanding [1].
From its early invention as a synthetic circuit by Gardner et al. [7], differ-
ent approaches for robust controlling of genetic toggle are proposed. Some
of them involve real time feedback control [12], auto regulation, noise [19],
the addition of some diffusible molecules like isopropyl-β-D-thiogalacto-
pyranoside (IPTG), and anhydrotetracycline (aTc) to control the promoter
activity as well. The search for a novel control parameter is still going on
and in some recent studies it was established that the limited availability of
cellular ingredients serving in the process of protein production can act as
a robust parameter in cell dynamics.
In the intermediate steps of protein production, majorly transcription
and translation, the genes collect resources for successful completion of
its expression from the cell. RNAP, transcription factor (TF), ribosome,
degradation machinery, etc. are various resources that the cell supplies to
the synthetic genetic circuit implemented in it or uses for its endogenous
gene functionalities. It is experimentally verified that the cell does not con-
tain these resources abundantly and depending upon the mode of opera-
tion, the availability of resources in different cells varies significantly. In the
translation process, the ribosome is considered to be the most important
Alternate Cell Fate in Bistable Genetic Switch 175
resource the gene circuit collects from the cell. Though the presence of this
essential cellular resource in all the prokaryotes and eukaryotes is a fact,
the amount of free ribosome in different living beings is different. Even for
different cells having different functionalities, they differ in availability of
free ribosomes. Like the pancreatic cells, in eukaryotes, they are dedicated
for most of the protein production over other cells and contain an unusu-
ally high number of ribosomes. On the contrary, the smooth endoplasmic
reticulum (SER) does not have ribosomes on its surface and thus does not
participate in protein production. This limited availability of the essen-
tial translational resource inside the cell significantly affects the ongoing
dynamics. In a low protein activity or for a demand of low resource for the
implemented synthetic construct, this limited availability may not affect
things, but in higher protein activity or for a larger resource demand for
the synthetic construct, unprecedented resource competition comes into
the picture. Different experimental and theoretical studies establish that
ribosome limitation, significantly modifying the circuit dynamics and at
a larger scale the system chooses the favorable state. In a recent study, it is
established that the protein production curve can be largely modified in
terms of sensitivity and amplification by controlling ribosome availabil-
ity and its distribution in the system [5]. Due to the nature of coupling
of different ongoing processes, the competition in outer motifs, which
seems unimportant in the study of motif interest, significantly modifies the
dynamic behavior and can ruin the entire system, resulting in some emer-
gent responses in output [4]. Not only limited to ribosomes, this competi-
tion can arise for RNAP, gene copy numbers, degradation machinery, and
many more. Yuriy Mileiko and team in a recent study have shown that gene
copy number variation brings a significant change in the dynamics of some
well-known motifs [13]. The effect of decoy binding is also established by
some recent papers [9, 11]. This trend of exploring the consequences of
resource availability and distribution in cell dynamics is new and scientists
are getting exciting novel mechanisms of system modelling from it.
Cellular decision-making, environmental sensing, and cell to cell com-
munication are three key processes underlying pattern formation and
development in microscopic and complex organisms. From a theoretical
point of view, though the cellular decision making seems to be reversible,
in practice most of the time this is irreversible due to some secondary
effects arising from the process. The biologically programmed cell death
(apoptis) or cell death in response to injury (lysis) is most the promising
reason here. Several approaches in determining cell fate like feedback-con-
trolled regulation, cell size, and growth rate dependency [10, 17] are estab-
lished in the recent past. The effect of noise, especially when a group of cells
is participating, statistically predicts the most probable result, as shown

in [1]. Though the presence of these various efforts of determining cell
fate transition, the effect of limited availability of resource or asymme-
try in resource affinity inside the cell and in the cell fate dynamics is not
explored much. This asymmetry serving as initial biasness has the capacity
to pre-pattern the cell fate.
In this paper, we have taken a simple genetic toggle motif and consid-
ered, along with the mutual repression of each other, that both the genes
are collecting resources (say ribosome here) from the same pool with dif-
ferent affinities. We also explore the condition for variability in gene copy
numbers for the pitchfork type of bifurcation here. Some of the major find-
ings of this paper are:
• Variability in resource affinity as well as gene copy number

introduces asymmetry in pitchfork bifurcation and pre-
patterns the cell fate. The greater the asymmetry, the more the
system pre-patterns itself in determining an alternate cell fate.
• Asymmetry in resource affinity regulates the range of bista-
bility and thus robustness of the switch in saddle node
bifurcation.
• Total availability of resource regulates the point of bifurca-
tion and region of interest in the system.
12.2 Model Formulation

Let us consider X and Y are the two proteins repressing each others pro-
moter activity, forming a toggle switch, as shown in Figure 12.1. The ham-
merhead symbols represent the repression here. We also consider both
these proteins are collecting resources from the same pool for their expres-
sion. We particularly focus on the pool of ribosome here, the essential
resource for the translation process, for the transcriptional complex to be
translated as a protein product.
Moving a step closer to cell dynamics, we consider that the ribosomes
are distributed over small several cytoplasmic compartments inside the
cell. Let T represent this local pool of resources available in the immediate
vicinity of the toggle switch. Thus, our consideration of the two partici-
pants of the concerned toggle switch collects resource ribosomes from the
pool T, which stands from its biological relevancy without any doubt.
The available mRNA pool for translation is represented by gx and gy for
protein X and Y, respectively.
X Y
resx resy
Resource T
(a) (b)
Steady states of protein
Steady states of protein
Bistable
region
Monostable Monostable
region region Monostable Bistable
region region
Bifurcation parameter Bifurcation parameter
(c) (d)
Figure 12.1 Model Motif. (a) Genetic toggle, two proteins X and Y mutually repress each
others promoter activity. Hammerhead symbol represents the repression here. Also, both
the participants of the toggle collect resources from the same pool T with the affinities resx
and resy respectively. (b) Schematic diagram of inter cellular competition determining cell
fate. Strength of repression drives the cell to a particular fate. (c) Saddle node bifurcation
in genetic toggle. A bistable region separates two monostable regions in phase space.
(d) Pitchfork bifurcation in genetic toggle. Monostable region switches to a bistable region
at bifurcation point.
The available mRNA pool gx and gy, collecting ribosomes from the pool
T, with affinities resx and resy, makes a ribosome bound complex cx and cy
which will be translated to protein X and Y at a rate of ∈x and ∈y, respec-
tively. This asymmetry in resource allocation is very insightful here. The
polycistronic mRNA pool in most of the bacterial organisms contains mul-
tiple ribosome binding sites (RBS) [3] and the rate of translation depends
on the rate of recruitment of ribosomes to this RBS, as well as on the rate
of translation initiation. The rate of ribosome recruitment also depends
upon many factors. Including all these rates, we generalize the resource
allocation or the protein production, which can have different rates as well.
As mentioned, from the total pool T, the ribosome bound complexes
are presented by cx and cy, thus further free pool of resource ribosome for
translation is estimated by (T − cx − cy).
The mutual repression is captured by a Michaelis–Menten type term in
our model and the hill function co-operativity n is taken as 2. The ODE
representing the scenario is given by Equation 12.1 below.
dc x
= res x (T − c x − c y ) g x − c x
dt
dX c x ∈x
= −X
dt 1 + Y n
(12.1)
dc y
= res y (T − c x − c y ) g y − c y
dt
dY c y ∈y
= −Y
dt 1 + X n
In a steady state, all the rates of change are equal to 0 and we investigated
the system.
12.3 Result Section

12.3.1 Pitchfork Bifurcation in Genetic Toggle
Pitchfork bifurcation occurs at specific equilibrium with perfect symmetry
conditions of the toggle system. For one variable dynamic system, several
studies are present, while for a two variable dynamic system, a few studies
in the recent past [2] investigated this phenomena in brief. For a conven-
tional pitchfork model, we take the production rates of the two proteins to
be equal, thus here we take ∈x, the production rate of X from its complex cx,
and, ∈y, the rate of production of Y from its complex cy is equal.
12.3.1.1 Resource Affinity Regulates the Symmetry

of Pitchfork Bifurcation
Following the conventional way of pitchfork bifurcation in a toggle sys-
tem, the protein production rates are equal and a detailed literature review
strongly supports our consideration of taking different resource affinity

values without any loss of generality. We find that the resource affinity
value regulates the symmetry of pitchfork bifurcation significantly. We
investigated the model for fixed values of n = 2, gx = gy = 5, T = 5, resy = 2,
and changing ∈x so that for every point ∈x = ∈y for four different values
of resx. When resx = resy, we find a beautiful symmetric pitchfork bifurca-
tion in the output, while more interestingly asymmetry in resource affinity
values destroys the symmetry in the output pitchfork with a significant
impression. Starting with a lower resource affinity for X, resx < resy, there is
a smooth transition of the pitchfork to a low state of the system, as shown
in Figure 12.2a, while a higher state is only accessible for a very large per-
turbation in the system. While starting from a higher resource affinity, resx
> resy for Figure 12.2d, the continuous accessible state is the higher pro-
duction state, while the lower production states are only accessible for a
large perturbation in the system. The same resource affinity of X and Y
7
10
Concentration of protein X
6 rex = 1 rex = 1.8

5 8
4 6
3 4
2
2
1
1 2 3 4 5 1 2 3 4 5
Rate of X production (Єx) Rate of X production (Єx)
(a) (b)
12 14
rex = 2
rex = 3
10 12
8 10
8
6
6
4
4
2 2
1 2 3 4 5 1 2 3 4 5
(c) (d)
Figure 12.2 Resource Affinity Regulates Symmetry of Pitchfork Bifurcation. Concentration

of protein X wrt the rate of X production ∈x plot. n = 2, resy = 2, gx = gy = 5, T = 5, for all the
plots. Resource affinity for X production, i.e., resx =1 for (a), resx=1.8 for (b), resx = 2 for (c),
and resx = 3 for (d).
(resx = resy = 2 for Figure 12.2c) results in a symmetric pitchfork. Also, it

is interesting to note that the larger the asymmetry, the higher the stability
of the chosen transitioned state and the lower the chances for its transi-
tion to another steady state even when perturbation is present from the
outside (comparing Figure 12.2a and Figure 12.2d with Figure 12.2c). So,
the results indicate to the conclusion that an initial asymmetry in resource
allocation pre-patterns the cells to a higher production regime or in a lower
production regime and determines the cell fate.
12.3.1.2 Availability of Total mRNA Pool Regulates the Symmetry

of Pitchfork Bifurcation
We find similar results with respect to the gene copy number available for
translation for a particular protein. The number of copies of a particular
gene present in the genotype is usually called the gene copy number. A
symmetry in the presence of gene copy number with symmetry in other
system parameters shows a perfect pitchfork in the output, while asymme-
try in the initial condition of gene copy number pre-patterns the system to
a higher or lower production state, as shown in Figure 12.3. It is interesting
to note that the greater the asymmetry in the initial gene copy number, the
greater the perturbation the system demands to transit from its continuous
steady state to the other.
12.3.1.3 Total Resource Availability Regulates the Point

of Bifurcation in the System
We find that total resource availability significantly regulates the bifurca-
tion point of the system, as shown in Figure 12.4. We investigate the model
for two fixed values of T, increasing the total resource available for transla-
tion and the position of bifurcation comes to a lower value of input signal.
Also, the range of steady states drastically changes.
12.3.2 Saddle Node Bifurcation in Genetic Toggle

The genetic toggle, most conventionally known as the genetic toggle switch,
is biologically most important for its on/off switch like behavior which
plays a significant role in determining cell fate. From its early invention,
researchers are deliberately searching for the ways of robust controlling of
the toggle switch. We find beautiful control on the toggle switch by con-
trolling resource distribution.
6 8
gx = 3 gx = 4.5
5 6
4
3 4
2 2
1
1 2 3 4 1 2 3 4
(a) (b)
10
8
gx = 5 8 gx = 6
6
6
4
4
2 2
1 2 3 4 1 2 3 4
(c) (d)
Figure 12.3 Availability of Total mRNA Pool Regulates Symmetry of Pitchfork Bifurcation
and Cell Fate Transition. Concentration of protein X vs. rate of X production ∈x plot. resx =
resy = 2, n = 2, T = 5, gy = 5 is fixed for all the plots and ∈x = ∈y. Gene copy numbers available
for production of X, i.e., gx =3 for (a), gx = 4.5 for (b), gx = 5 for (c), and gx = 6 for (d).
7
T=5 14
6 T=10
12
5 10
4 8
3 6
2 4
1 2
0.5 1.0 1.5 2.0 2.5 3.0 0.5 1.0 1.5 2.0 2.5 3.0
(a) (b)
Figure 12.4 Total Resource T Availability Regulates Point of Pitchfork Bifurcation in

Genetic Toggle. Concentration of protein X vs. rate of X production ∈x plot. n = 2, gx =
gy = 5, ∈x = ∈y, resx = resy = 2 for both the plots. T = 5 for (a) plot and T = 10 for (b) plot.
12.3.2.1 Resource Distribution Regulates the Point of Bifurcation

in Toggle Switch
We find the resource distribution significantly regulates the point of bifur-
cation in the saddle node of genetic toggle as well, as shown in Figure 12.5.
We investigate the system for 3 different values of resx, keeping all other
parameters fixed at resy = 2, n = 2, gx = gy = 5, T = 5, and ∈y = 2 and plot the
concentration of protein X with respect to the activator of X production ∈x.
Considering the blue line (continuous and dashed) primarily, which shows
the scenario when resx = resy = 2 and that resource allocation for X to Y is
the same, we find a change in resx shifts the curve left to the green curve (or
right to the red curve) for a higher affinity for resources to X than Y, resx =
3, > resy = 2 (for a lower affinity for resource to X than Y, resx = 1, < resy = 2).
12.3.2.2 Region of Interest in Toggle Switch is Significantly

Regulated by Resource Allocation
For a saddle node bifurcation, the most interesting region is the range of
input signal for which the output protein concentration attains two dras-
tically different concentrations depending upon the mode of forward or
backward operation. When investigated in a bifurcation diagram, a set
of stable equilibrium points is separated by unstable equilibrium points,
the system cannot achieve physically. From Figure 12.5, we also get that
the range of interest increases (or decreases) for X with a lower resource
allocation resx = 1 < resy = 2 (for X getting higher resource than Y, resx =
1.6
10 1.5
Range of bistable region (Єx)
1.4
8
1.3
6 resx = 1 1.2
resx = 2 1.1
4 resx = 3
1
0.9
2
0.8
0 0.7
0 2 4 6 8 5 5.5 6 6.5 7 7.5 8 8.5 9 5.5 10
Rate of X production (Єx) Available Total resource T
(a) (b)
Figure 12.5 (a). Rate of X production ∈x vs. concentration of protein X plot in case
of saddle node bifurcation in genetic toggle. resy = 2, gx = gy = 5, T = 5, n = 2, ∈y = 2.
(b). Phase plot for range of bistable region ∈x, the rate of X production from its complex
vs. total resource availability T. resx = resy = 1, n = 2, gx = gy = 5, ∈y = 1.
3 > resy =2). It is interesting to note that lower resource availability to X

than Y stabilises the bifurcation curve for larger fluctuation. This is not
only giving us an opportunity for robust controlling of the system, but also
signifies that initial biasing of the system towards the resource significantly
modifies cell fate in terms of stability.
12.3.2.3 Total Resource Availability T Regulates Saddle Node

Bifurcation Curve
Along with the variation in resource affinity, the availability of the total
nutrients here significantly regulates the saddle node bifurcation curve,
as shown in Figure 12.5b. The regulation is quite positive here though.
Greater resources stabilize the system for a larger range of bistability and
greater switch robustness, while with less availability of resources the
switch response is not stable and the steady states can alter even for low
fluctuation in the system.
12.4 Conclusion
Cellular decision making is a fundamental biological phenomenon by
which a cell opts the different states prior to environmental conditions,
leading to asymmetric cell differentiation. The underlying reason behind
this is still not entirely explored. We take a simple genetic motif, genetic
toggle here, and show that resource affinity asymmetry of the toggle par-
ticipants, both for the saddle node and in pitchfork bifurcation, signifi-
cantly biases the cell fate. This mutually repressing motif is very common
in nature [6, 14], where in output the system shows patterning by choosing
one cell fate over another. The availability of total resource pool signifi-
cantly regulates the bifurcation point in the motif. In the case of a synthetic
circuit, this resource limitation is very true because the gene circuit imple-
mented in the host entirely depends upon the host’s resource for its expres-
sion and in the case of a cell, the limitation in cellular resource ribosome
is a major factor, indicating our findings to be true also. We also investi-
gated the effect of gene copy number in the case of pitchfork bifurcation,
indicating an initial asymmetry that biases the cell fate to lower or higher
production states accordingly.
Here, it is important to note that our entire consideration is valid for a
low growth state of the system. The effect of growth rate on cell dynamics
is well established [17]. Overexpression of endogenous genes, or adding
some synthetic construct in the cell, destabilizes the resource distribution

and makes the growth rate smaller, while growth causing dilution enhances
protein degradation. So, some researchers point out that growth is a sig-
nificantly regulatory parameter in every cellular phenomenon, but some
experimental results also pointed out that these effects only depend upon
experimental conditions, causing some momentary changes in dynam-
ics. Our study mostly follows the experimental situation [8, 15] when the
growth rate is low and competition effect is significant.
A perfect noise free environment in the cell is impracticable. For a single
cell, though the consideration does not violate the reality, working with
a group of cells, the predictability can vary significantly. The addition of
noise in existing dynamics will give a result close to reality. Also, the lim-
itation of other cellular nutrients in the way of gene expression can regulate
the alternate protein production, stabilizing one state over other, regulat-
ing cell fate. In the future, we would like to extend our work for a complete
scenario including transcriptional, translational, and degradation machin-
ery competition in a noisy cell environment.
Acknowledgement
PC and SG acknowledge the support by DST-INSPIRE, India, vide sanc-
tion Letter No. DST/INSPIRE/04/2017/002765 dated- 13.03.2019.
References
1. G. Balázsi, A. van Oudenaarden, and J. J. Collins. Cellular decision mak-
ing and biological noise: from microbes to mammals. Cell, 144(6):910–925,
2011.
2. I. Bose and S. Ghosh. Bifurcation and criticality. Journal of Statistical
Mechanics: Theory and Experiment, 2019(4):043403, 2019.
3. D. H. Burkhardt, S. Rouskin, Y. Zhang, G.-W. Li, J. S. Weissman, and C. A.
Gross. Operon mrnas are organized into orf-centric structures that predict
translation efficiency. Elife, 6:e22037, 2017.
4. P. Chakraborty and S. Ghosh. Emergent correlations in gene expression
dynamics as foot-prints of resource competition. The European Physical
Journal E, 44(10):1–12, 2021.
5. P. Chakraborty and S. Ghosh. Emergent regulatory response and shift of half
induction point under resource competition in genetic circuits. arXiv pre-
print arXiv:2112.04985, 2021.
6. E. Clark. Dynamic patterning by the drosophila pair-rule network reconciles

long-germ and short-germ segmentation. PLoS biology, 15(9):e2002439,
2017.
7. T. S. Gardner, C. R. Cantor, and J. J. Collins. Construction of a genetic toggle
switch in escherichia coli. Nature, 403(6767):339–342, 2000.
8. A. Gyorgy, J. I. Jiménez, J. Yazbek, H.-H. Huang, H. Chung, R. Weiss, and D.
Del Vecchio. Isocost lines describe the cellular economy of genetic circuits.
Biophysical journal, 109(3):639–646, 2015.
9. S. Jayanthi, K. S. Nilgiriwala, and D. Del Vecchio. Retroactivity controls the
temporal dynamics of gene transcription. ACS synthetic biology, 2(8):431–
441, 2013.
10. S. Klumpp, Z. Zhang, and T. Hwa. Growth rate-dependent global effects on
gene expression in bacteria. Cell, 139(7):1366–1375, 2009.
11. T.-H. Lee and N. Maheshri. A regulatory role for repeated decoy transcrip-
tion factor binding sites in target gene expression. Molecular systems biol-
ogy, 8(1):576, 2012.
12. J.-B. Lugagne, S. S. Carrillo, M. Kirch, A. Köhler, G. Batt, and P. Hersen.
Balancing a genetic toggle switch by real-time feedback control and periodic
forcing. Nature communications, 8(1):1–8, 2017.
13. Y. Mileyko, R. I. Joh, and J. S. Weitz. Small-scale copy number variation
and large- scale changes in gene expression. Proceedings of the National
Academy of Sciences, 105(43):16659–16664, 2008.
14. Y. Saka and J. C. Smith. A mechanism for the sharp transition of morphogen
gradient interpretation in xenopus. BMC developmental biology, 7(1):1–9,
2007.
15. I. Shachrai, A. Zaslaver, U. Alon, and E. Dekel. Cost of unneeded proteins in
e. coli is reduced after several generations in exponential growth. Molecular
cell, 38(5):758–767, 2010.
16. M. Strasser, F. J. Theis, and C. Marr. Stability and multiattractor dynamics
of a toggle switch based on a two-stage model of stochastic gene expression.
Biophysical journal, 102(1):19–29, 2012.
17. C. Tan, P. Marguet, and L. You. Emergent bistability by a growth-modulating
positive feedback circuit. Nature chemical biology, 5(11):842–848, 2009.
18. T. Tian and K. Burrage. Bistability and switching in the lysis/lysogeny genetic
regulatory network of bacteriophage λ. Journal of Theoretical Biology,
227(2):229–237, 2004.
19. J. Wang, J. Zhang, Z. Yuan, and T. Zhou. Noise-induced switches in network
systems of the genetic toggle switch. BMC systems biology, 1(1):1–14, 2007.
13
A Hybrid Approach to
Ontology Evaluation
Aastha Mishra* and Preetvanti Singh
Department of Physics & Computer Science, Dayalbagh Education Institute, Agra,

Uttar Pradesh, India
Abstract
In last few decades, researchers have been motivated to facilitate ontology in sev-
eral fields. In medical science, ontology is used to describe the theory of medical
vocabularies and the correlation shared among them, thus permitting the sharing
of medical knowledge. Definition based on ontology of disease allows each class of
disease to be classified in a formalized structure singularly and along with the dis-
cussion of ontological realism for the treatment and diagnosis of disease. This paper
uses a hybrid approach to evaluate ontology for epilepsy disease. A multi-criteria
decision making (MCDM) method is used to decide the best ontology provided by
Bio-portal to select the best suitable characteristics for epilepsy disease.
Keywords: Ontology, ontology evaluation, AHP, epilepsy, MCDM
13.1 Introduction
Nowadays, the single resource most critical for top management in an
organization to sustain competitive advantage is knowledge. An orga-
nization can have competitive edge over its competitors by building an
excellent process to manage knowledge. Ontology is a static knowledge
representation method that includes definitions of basic concepts in a
domain and relations among them which are machine interpretable. The
information retrieval quality is improved from a keyword-based retrieval
to knowledge-based search. Users should have a way for assessing ontolo-
gies and deciding which one fits their requirements to best face a multitude
*Corresponding author: aasthamishra006@yahoo.in
187
of ontologies. The main benefit of developing an ontology is that it enables

sharing common understanding of the information structure and efficient
reuse of domain knowledge.
Ontologies are an elementary data structure for knowledge conceptual-
ization, but many ontologies are built for conceptualizing the same body
of knowledge. Ontology selection is thus an important issue that must
be addressed to access these and decide which one fits the requirements
of the user best. Evaluating characteristics of ontologies has also become
necessary because of the increasing number of candidates for reuse in a
domain and complexities of ontologies. Various approaches for ontol-
ogy selection have been considered in literature [6, 7, 9, 11, 20, 23]. A
multi-criteria decision-making approach to ontology selection deals with
the problem of selecting the ontology best suited to the needs of the deci-
sion maker.
Multi-criteria decision making (MCDM) assigns the selection of the
choice of a best alternative from several available options in a decision,
subject to several vague or concrete criteria or attributes. It helps peo-
ple make decisions according to their choices in cases where more than
one conflicting criterion exists. Analytical Hierarchy Process (AHP), an
MCDM process, is a structured decision-making tool for quantifying the
weights of decision criteria. Experience of experts is utilized to estimate
the relative magnitude of parameters through pairwise comparisons. This
paper presents a multi-criteria decision making approach based method
to evaluate ontologies and select the best ontology in the medical domain.
13.2 Background
Methods have been developed to evaluate and select the best ontology. [16]
explored the gain utilization of Semantic Web technologies in the domain
of recruitment and developed an ontology based recruitment process. [17]
used ontology evaluation techniques for agent cooperation to measure the
quality of ontology. A criteria selection framework was proposed by [10]
for guiding the selection of suitable criteria for various levels of ontology
evaluation. [5] presented a method for ontology evaluation based on the
goal, question, and metric approach for empirical evaluation. [15] pro-
posed a method to integrate ontologies to select and recommend adapted
internship seekers. [22] presented metrics, approaches, and other similar
aspects of ontology evaluation in a concise manner.
[12] presented a scalable data-driven framework for ontology evaluation,
targeting Big Data scenarios and use cases. [21] presented the outstanding
A Hybrid Approach to Ontology Evaluation 189
contribution to ontology evaluation by considering social and community

related themes. [2] organized four-categorical schemes to evaluate ontol-
ogy in the existing literature. 200 ontology samples are considered, which
are taken from the National Center for Biomedical Ontology (NCBO) Bio
Portal. [4] evaluated the U Ontology by ontology evaluation methodology
to evaluate the quality of the developed ontology. [11] discussed already
present evaluation metrics that support the ontology evaluation process
for offering guidance useful to knowledge handling, representation, and
conceptualization. [13] explored different metrics used for the evaluation
of quality of ontology from different dimensions for ontology evaluation.
An approach to evaluate quality of reused parts was proposed by [24]. The
model represented evaluation information with semantic properties.
[1] presented OnToology, a web-based application to manage ontol-
ogy engineering support activities. [3] presented the 5 ontologist’s results,
revealing high system usability of OntoKeeper and use-cases. [8] evaluated
a set of consistent and objective ontology structural metrics. Ontology
repositories used for evaluation have been used as corpora. [14] developed
a decision support system for manufacturing process selection based on
ontology-enabled case-based reasoning. From the review, it was observed
that MCDM techniques are used to provide an efficient way to evaluate
and select ontologies. This paper develops a method for evaluating and
selecting ontologies in the medical domain based on the characteristics of
a disease.
13.3 The Developed OntoEva Method

Ontology selection is the method of identifying the ontologies or ontology
models that best suit the requirements of a decision maker. The precondi-
tion of ontology selection lies in evaluating all considered ontologies on
the basis of certain criteria. The same body of knowledge is conceptualized
by different ontologies, so the method developed in this paper will enable a
user to select the one that best suits the requirements of the decision maker.
The OntoEva method was developed to evaluate and select the ontolo-
gies using AHP. The use of AHP will enable a decision maker in evaluating
the ontologies based one some prioritized criteria. The proposed method
determines the most suitable ontology by considering a set of characteristics
and evaluation criteria. Weights are generated for each evaluation criteria
according to pairwise comparison of the criteria and the criterion with the
highest weight is considered as the most important criteria. Computation
of the weights is guided by the experience of decision makers. OntoEva
first identifies the characteristics of ontologies for a domain and computes

weights for each criterion. The class mapping is then observed for priori-
tized factors, which helps in determining the best ontology for the domain.
The methodology steps are as follows:
Step 1: Identify the information need.

Step 1.1: Identify the domain and the information need.
Step 1.2: Identify ontology structures that satisfy the infor-
mation need.
Step 2: Specify the selection criteria. These criteria are similar to
ontology popularity, topic coverage, or ontology structure.
Step 3: Identify n characteristics of ontology for a domain.
Step 4: Categorize these characteristics in m clusters.
Step 5: Construct a pairwise comparison matrix for each of
the m clusters using the 9-point scale [18] and calculate
(Inw) importance weights for the characteristics (criteria).
Table 13.1 shows the 9-point scale is change to AHP pair-
wise comparison table.
Table 13.1 9-point scale [18].
Value of ajk Interpretation
1 Same importance
3 Average importance of one over another
5 Important or strong importance
7 Extreme importance
9 Very extreme importance
2,4,6,8 Middle values between the two adjacent judgments
Step 6: For the prioritized characteristics in each cluster, per-

form class mappings.
Step 7: Based on class mapping, compute importance weights
for the considered ontologies.
13.4 Ontology Selection for Epilepsy Disorder

Epilepsy is a neurological disorder that affects people of all ages. According
to [19], of the total population approximately 2% suffers from epilepsy,
which ranks it as the second most common neurologic disorder. A person
with epilepsy experiences symptoms like sensations, irritability, headache,

depression, ‘funny feeling’, abnormal behavior, confusion, and sometimes
loss of consciousness. The main reason of the suffering is lack of awareness
about the disease. Thus, it becomes necessary that a commoner is provided
a method to efficiently diagnose the disease at early stages. The three ontol-
ogy structures considered in this study for this goal are:
• Extended Syndromic Surveillance Ontology (ESSO): To facil-

itate the mining of free-text clinical documents, an open-
source terminological ontology is designed in English. It
consists of epilepsy syndromes, seizure types, and data ele-
ments associated with them.
• Epilepsy and Seizure Ontology (EPSO): To support epilepsy
focused informatics, tools this ontology are developed for
patient care and clinical research.
• Epilepsy Ontology (EPILONT): It is ontology about the
epilepsy domain and epileptic seizure based on the diag-
nosis proposed by the ILAE (International League Against
Epilepsy).
The fundamental properties considered for the selection process are as

follows.
13.4.1 Accuracy
To obtain higher accuracy, correct definitions and descriptions of classes,
properties, and individuals which clearly define the domain are required,
i.e., the considered ontologies should specify epilepsy disease.
13.4.2 Adaptability
Adaptability defines how long the ontology predicts its uses. Ontologies for
epilepsy are originally designed to describe the criteria on epilepsy disease
and its vocabulary also allows formalizing symptoms of all kinds and dif-
ferentiating between diseases.
13.4.3 Clarity
Clarity calculates how productively the ontology can communicate with
the deliberate meaning of the given terms. The name of the ontology
should clearly explain the content and its function.
13.4.4 Completeness
For completeness, the domain of interest and the thickness and rich-
ness of the ontology should be properly covered. In order to identify the
disorder, the list of all relevant characteristics is provided by the three
ontologies.
13.4.5 Conciseness
Conciseness is the evaluation criteria that states if the ontology includes
irrelevant elements regarding to the domain to be covered. For example,
ontology about epilepsy disease may take an important view on what the
disease actually is. It is not important to state if a person suffering from
other disease and any related information about that.
13.4.6 Consistency
Consistency explains that the ontology should not involve or allow for any
discrepancies, for example: Confused being Confused Memory is the one
of the symptoms of epilepsy, but having a logical axiom, for example: call-
ing confused a mental state will contradict the statement.
13.4.7 Organizational Fitness

Organizational fitness involves various factors deciding the ease of how
ontology can be deployed within an organization. A hospital may decide
that all used ontologies align to the Bio-portal. This will help the organiza-
tion in reducing costs when integrating data from different sources to align
the ontologies.
The three ontologies, EPSO, EPILONT, and ESSO, fulfill all the above
mentioned properties and are suitable for selecting the best ontology for
epilepsy disease.
After selecting the ontologies, the topic coverage of these ontologies was
analyzed from the Bio-portal, as given in Tables 13.2, 13.3, and 13.4. The
characteristics analyzed were the ontology metric classes, properties, indi-
viduals, maximum depth, and maximum number of children.
In order to evaluate ontologies for epilepsy disease, the characteristics
of epilepsy are selected after having discussions with doctors. Considering
their opinions, the selected criteria are clustered under m=4 groups, as
shown in Table 13.5, and were decomposed into hierarchy as illustrated in
Figure 13.1.
Table 13.2 EPSO ontology.

Classes 1357
Individuals 2
Properties 29
Maximum depth 17
Maximum number of children 146
Table 13.3 EPILONT ontology.

Classes 138
Individuals 0
Properties 10
Maximum depth 4
Table 13.4 ESSO ontology.

Classes 2705
Individuals 0
Properties 166
Maximum depth 12
For each of the clusters, a pairwise comparison matrix is developed

(Table 13.6). These matrices were consistent, as the CI and CR values were
< 0.1.
The computed weight for each characteristic is given in Table 13.7.
From the computed weights it can be seen that the focal type (A3) is the
highest rated characteristic, AED (B2) is the highest weighted treatment,
Head Trauma (C1) is the highest rated cause of epilepsy, and Confused
Memory (D2) is the most noticeable symptom due to the highest weight
value.
Table 13.5 Selected criteria.

Cluster Description Criteria
[A] Types Epilepsy is diagnosed in A1 – Absence
people when they have A2 – Atomic
two or more seizures. A3 – Focal
A4 – Generalized
A5 – Tonic
[B] Treatment The goal of treatment in B1 – Ketonic Diet
patients suffering from B2 – AED
epileptic seizures is to B3 – Physical Exercise
achieve a seizure-free B4 – Surgery
status without any side B5 – Nerve Stimulation
effects.
[C] Causes The causes can be complex C1 – Head Trauma
and sometimes hard to C2 – Brain Condition
identify. C3 – Prenatal Injury
C4 – Alcohol Consumption
C5 – Genetics
[D] Symptoms Symptoms differ from D1 – High Fever
person to person and D2 – Confused Memory
according to the type of D3 – Fainting
seizure. D4 – Narcolepsy
D5 – Cataplexy
D6 – Panic Attack
D7 – Breathing Difficulty
EPILEPSY
Types Treatment Causes Symptoms
A1 A2 A3 A4 A5 B1 B2 B3 B4 B5 C1 C2 C3 C4 C5 D1 D2 D3 D4 D5 D6 D7
EPSO ESSO EPILONT
Figure 13.1 Hierarchy for epilepsy disorder.

Table 13.6 Pairwise comparison matrix for clusters.

Cluster: Type
A1 A2 A3 A4 A5
A1 1.000 0.333 0.200 0.333 0.333
A2 3.000 1.000 0.250 0.333 0.500
A3 5.000 4.000 1.000 4.000 4.000
A4 3.000 3.000 0.250 1.000 0.500
A5 3.000 2.000 0.250 2.000 1.000
Cluster: Treatment
B1 B2 B3 B4 B5
B1 1.000 0.250 3.000 0.500 0.500
B2 4.000 1.000 3.000 2.000 2.000
B3 0.333 0.333 1.000 0.500 0.333
B4 2.000 0.500 2.000 1.000 2.000
B5 2.000 0.500 3.000 0.500 1.000
Cluster: Causes
C1 C2 C3 C4 C5
C1 1.000 2.000 2.000 3.000 5.000
C2 0.500 1.000 2.000 2.000 3.000
C3 0.500 0.500 2.000 2.000 3.000
C4 0.333 0.500 1.000 1.000 2.000
C5 0.200 0.333 0.500 0.500 1.000
Cluster: Symptoms
D1 D2 D3 D4 D5 D6 D7
D1 1.000 0.333 0.500 2.000 3.000 4.000 2.000
D2 3.000 1.000 2.000 2.000 3.000 4.000 5.000
D3 2.000 0.500 1.000 3.000 2.000 4.000 3.000
(Continued)
Table 13.6 Pairwise comparison matrix for clusters. (Continued)

D4 0.500 0.500 0.333 1.000 0.500 3.000 2.000
D5 0.333 0.333 0.500 2.000 1.000 3.000 2.000
D6 0.250 0.250 0.250 0.333 0.333 1.000 3.000
D7 0.500 0.200 0.333 0.500 0.500 0.333 1.000
After identifying the prioritized characteristics for each cluster, each of

these characteristics was searched in each of the three ontologies.
For Cluster I, Type, Focal was present in the class mapping.
These diagrams illustrate the presence of characteristic class in each of
the ontologies.
EPSO Ontology (7-mapping)
Continuant
Dependent Continuant
Diease
Epilepsy
Focal Epilepsy
ESSO Ontology (16-mapping)
Classification_System
Blume_2001_ILAE_Glossary
Blume_2001_I_General_Terms
Blume_2001_I_5.0_Focal
Focal
Table 13.7 Computed weights.

Cluster Criteria Inw
A A1 0.058
A2 0.103
A3 0.488
A4 0.162
A5 0.191
B B1 0.128
B2 0.373
B3 0.081
B4 0.230
B5 0.188
C C1 0.382
C2 0.246
C3 0.185
C4 0.118
C5 0.068
D D1 0.166
D2 0.298
D3 0.213
D4 0.098
D5 0.113
D6 0.060
D7 0.052
EPILONT Ontology (0-mapping)

Seizure Types
Crises Continuas
Focal Status Epilepticus
According to above data, the focal type of epilepsy exists in all the three
ontologies. The focal type of epilepsy is the most prominent type among
all the epilepsy types. Most patients suffering from epilepsy suffer from
this type of the disease. In EPSO type presents with 7-class mapping, ESSO
presents a focal class with 16-class mapping, and EPILONT presents with
0-mappping. Considering the above data, pairwise comparison matrixes
are generated and Ontology Weights (Ow) are computed using the Eigen-
vector approach.
EPSO EPILONT ESSO

EPSO 1.000 5.000 0.333
EPILONT 1.000 0.143
ESSO 1.000
The results are consistent, as the values of CI and CR were less than 0.1.
The same procedure is performed for all other selected criteria.
For Cluster II, Treatment, the class mapping of each of the considered
Epilepsy Ontology was observed. The Treatment AED is present in class
mapping.
EPILONT (0 – mapping)
Sindromes Epilepticos
Crises que nao obrigam neccessariamente ao diagnostic de Epilepsia
Drug or other chemically - induced seizures

EPSO (7- mapping)

Continuant
Independent Continuant
Clinical Drug Component
DrugBrandName
ESSO (1- mapping)

Treatment
Medication
Medication_Name
In EPSO, the AED class is present as a clinical drug component and its sub-
class drug brand name with 7-class mapping, ESSO consist of Medication_
Name class with 1-class mapping, and EPILONT consists of 0-mappping.
EPSO EPILONT ESSO

EPSO 1.000 5.000 3.000
EPILONT 1.000 0.500
ESSO 1.000
For Cluster III, Causes, Head Trauma is present in class mapping as follows:
EPILONT – not found
EPSO (18 – mapping)

Continuant
Dependent Continuant
Etiology
Structural Cause
Selerosis
Traumatic Brain Injury

ESSO (20 – mapping)

Hidden Terms
Buchhalter Meeting
Head Injury
Traumatic Brain Injury
In EPSO Ontology, Traumatic Brain Injury class is present with 20-class

mapping, ESSO consists of the same class name with 18-class mapping,
and in EPILONT it is not present.
EPSO ESSO
EPSO 1.000 1.000
ESSO 1.000
For Cluster IV Symptoms, Confused Memory is present in class mapping:
EPILONT (71- mapping)
General Concepts
Oncology mapping
Syncope Symptom Ontology
Neurological and Physiological Symptom
Confusion
EPSO (0-mapping)
Occurrent
Processual Entity
Bodily Process
Physical Process
Physical Pathological Process
Seizure Feature
Seizure Consciousness State

ESSO – Not Found

In EPSO Ontology, class seizure consciousness state is present with 0-class
mapping, ESSO does not have that class, and in EPILONT it consists of a
Confusion class with 71-class mapping.
EPSO EPILONT
EPSO 1.000 0.250
EPILONT 1.000
13.5 Results
On the basis of above analysis, the computed weights are shown in Table
13.8.
Finally, the average weights of each of the 3 ontologies are computed:
EPSO = 0.279 + 0.648 + 0.500+0.200 = 1.627/ 4 = 0.406
EPILONT = 0.072+0.122+ 0.500+0.800 =1.494/4 = 0.373
ESSO = 0.649 + 0.230+0+0 = 0.879/4 = 0.219
It can be seen that EPSO is the suitable ontology for diagnosing epilepsy
disease.
13.6 Comparison of Ontologies

When these ontologies are compared from the Bio-portal platform on the
basis of ontology metrics like classes, properties, individuals, maximum
depth, and maximum number of children, the most suited ontology is
ESSO.
However, from the computed weights, as shown in Table 13.8, EPSO
was considered as the best ontology for epilepsy disease. The result was
accepted by neurologists, as the EPSO provides all the required informa-
tion a patient needs and it contains all the classes that best defined the
epilepsy disease.
Table 13.8 Computed ontology importance weights.

Types – Focal
Ontology OW
EPSO 0.279
EPILONT 0.072
ESSO 0.649
Treatment – AED
Ontology OW
EPSO 0.648
EPILONT 0.122
ESSO 0.230
Causes – Head Trauma
Ontology OW
EPSO 0.500
EPILONT 0.500
ESSO 0.000
Symptoms – Confused Memory
Ontology OW
EPSO 0.200
EPILONT 0.800
ESSO 0.000
13.7 Conclusion
This paper presents a hybrid method for ontology selection. The method
uses AHP technique to compute weights for determining the best suited
ontology for diagnosing a disease. The study demonstrates its use for diag-
nosing epilepsy. Three ontologies, EPSO, ESSO, and EPILONT are con-
sidered for the study. Focal, AED, Head Trauma, and Confused Memory
were evaluated as the important for determining the best ontology among
these three.
For future reference, a more improved ontology for epilepsy disease can
be developed. Ontology Evaluation and Selection can be done using other
decision making techniques.
References
1. Alobaid, A., Garijo, D., Poveda-Villalón, M., Santana-Perez, I., Fernández-
Izquierdo, A., & Corcho, O. (2019). Automating ontology engineering sup-
port activities with OnToology. Journal of Web Semantics, 57, 100472.
2. Amith, M., He, Z., Bian, J., Lossio-Ventura, J. A., & Tao, C. (2018). Assessing
the practice of biomedical ontology evaluation: Gaps and opportuni-
ties. Journal of biomedical informatics, 80, 1-13.
3. Amith, M., Manion, F., Liang, C., Harris, M., Wang, D., He, Y., & Tao, C.
(2019). Architecture and usability of OntoKeeper, an ontology evaluation
tool. BMC medical informatics and decision making, 19(4), 152.
4. Ashraf, J., Hussain, O. K., Hussain, F. K., & Chang, E. J. (2018). Representation
Phase: Ontology Usage Ontology (U Ontology). In Measuring and Analysing
the Use of Ontologies (pp. 171-203). Springer, Cham.
5. Bandeira, J., Bittencourt, I. I., Espinheira, P., & Isotani, S. (2016). FOCA: A
methodology for ontology evaluation. ArXiv preprint arXiv:1612.03353.
6. Brank, J., Grobelnik, M., & Mladenic, D. (2005, October). A survey of ontol-
ogy evaluation techniques. In Proceedings of the conference on data min-
ing and data warehouses (SiKDD 2005) (pp. 166-170). Citeseer Ljubljana,
Slovenia.
7. Fonou-Dombeu, J. V. (2019, June). A Comparative Application of Multi-
criteria Decision Making in Ontology Ranking. In International Conference
on Business Information Systems (pp. 55-69). Springer, Cham.
8. Franco, M., Vivo, J. M., Quesada-Martínez, M., Duque-Ramos, A., &
Fernández-Breis, J. T. (2019). Evaluation of ontology structural metrics
based on public repository data. Briefings in bioinformatics.
9. Haghighi, P. D., Burstein, F., Zaslavsky, A., & Arbon, P. (2013). Development and
evaluation of ontology for intelligent decision support in medical emergency
management for mass gatherings. Decision Support Systems, 54(2), 1192-1204.
10. Hooi, Y. K., Hassan, M. F., & Shariff, A. M. (2015, May). Ontology evalua-
tion—A criteria selection framework. In 2015 International Symposium on
Mathematical Sciences and Computing Research (iSMSC) (pp. 298-303).
IEEE.
11. Jarosław, W. (2018). An Attempt to Knowledge Conceptualization of Methods
and Tools Supporting Ontology Evaluation Process. Procedia computer sci-
ence, 126, 2238-2247.
12. Knoell, D., Atzmueller, M., Rieder, C., & Scherer, K. P. (2017). A Scalable
Framework for Data-Driven Ontology Evaluation. In WM (pp. 97-106).
13. Lourdusamy, R., & John, A. (2018, January). A review on metrics for ontol-
ogy evaluation. In 2018 2nd International Conference on Inventive Systems
and Control (ICISC) (pp. 1415-1421). IEEE.
14. Mabkhot, M. M., Al-Samhan, A. M., & Hidri, L. (2019). An Ontology-
Enabled Case-Based Reasoning Decision Support System for Manufacturing
Process Selection. Advances in Materials Science and Engineering, 2019.
15. M’Baya, A., Laval, J., Moalla, N., Ouzrout, Y., & Bouras, A. (2016, December).
Ontology based multi criteria recomanded system to guide internship assign-
ment process. 3rd IEEE International Conference on Computational Science
and Computational Intelligence (CSCI 2016), Dec 2016, Las Vegas, United
States. ffhal-01788507f
16. Mochol, M., Oldakowski, R., & Heese, R. (2004). Ontology based recruit-
ment process. Informatik 2004, Informatik verbindet, Band 2, Beiträge der
34. Jahrestagung der Gesellschaft für Informatik eV (GI).
17. Plinere, D., & Borisov, A. (2014). Evaluation of the Ontological Knowledge
Model/Ontoloģiskā zināšanu modeļa novērtēšana/Оценка качества
онтологической модели знаний. Information Technology and Management
Science, 17(1), 81-85.
18. Saaty, T. L. (1988). What is the analytic hierarchy process? In Mathematical
models for decision support (pp. 109-121). Springer, Berlin, Heidelberg.
19. Stafstrom, C. E. (1998). The pathophysiology of epileptic seizures: a primer
for pediatricians. Pediatrics in review, 19(10), 342.
20. Sun, L., Ma, J., Zhang, Y., Dong, H., & Hussain, F. K. (2016). Cloud-FuSeR:
Fuzzy ontology and MCDM based cloud service selection. Future Generation
Computer Systems, 57, 42-55.
21. Talebpour, M., Sykora, M., & Jackson, T. (2017, November). Social and
Community Related Themes in Ontology Evaluation: Findings from an
Interview Study. In International Joint Conference on Knowledge Discovery,
Knowledge Engineering, and Knowledge Management (pp. 320-336).
Springer, Cham.
22. Verma, A. (2016, August). An abstract framework for ontology evalua-
tion. In 2016 International Conference on Data Science and Engineering
(ICDSE) (pp. 1-6). IEEE.
23. Vrandečić, D. (2009). Ontology evaluation. In Handbook on ontologies
(pp. 293-313). Springer, Berlin, Heidelberg.
24. Xu, F. X., Liu, X. H., Chen, W., Zhou, C., & Cao, B. W. (2018). An Ontology
and AHP Based Quality Evaluation Approach for Reuse Parts of End-of-Life
Construction Machinery. Mathematical Problems in Engineering, 2018.
14
Smart Health Care Waste
Segregation and Safe Disposal
R.M. Bommi1*, Sami Venkata Sai Rajeev2, Sarvepalli Navya2, Veluru Sai Teja2
and Uppala Supriya2
Saveetha School of Engineering, SIMATS, Chennai, India

1
2
Department of CSE, Chennai Institute of Technology, Chennai, India
Abstract
Medical waste disposal has been a big issue due to an exponentially growing pop-
ulation and the COVID-19 pandemic. Increased waste generation per person
has resulted from urbanization, industrialization, and economic development.
Substandard medical waste separation at the site of origin might have a cascad-
ing effect on the environment, putting humans, wildlife, soil, and water bodies at
risk. If hazardous airborne pollutants are not effectively controlled, separated, and
burned by on-site or off-site incineration, environmental concerns linked with
inadequate clinical waste may pollute the air we breathe. This paper proposes an
IoT based smart healthcare waste segregator which segregates the waste into five
kinds. The sensors detect the type of waste and the waste gets disposed into the
smart bins accordingly. Using artificial intelligence, the status of filling of the bin is
indicated through LEDs. When the bin reaches the maximum-level, an alert mes-
sage is sent to the municipal authorities. The filled waste gets wrapped automati-
cally. The wastes which need to be incinerated is burnt in the incinerator chamber
available in the system. Therefore, this system will capably make the environment
smart, clean, and safe.
Keywords: Sensors, IoT, artificial intelligence, incinerator, garbage, smart bins
*Corresponding author: rmbommi@gmail.com
205
14.1 Introduction
The Internet plays an important role in today’s world by linking com-
puters to the worldwide web (www), which permits users to access data
from anywhere in the world [1]. The Internet of Things (IoT) refers to
things that are connected to the internet and can often be managed from
there [2]. Garbage is described as solid substances generated as a result
of human activities that are removed from the system [3] because they
are no longer useful in the respective economic, biomedical, or technical
method. In a wider context, solid waste refers to all products that are
used in the home, industry, or agriculture. Municipal solid waste (MSM)
is described as waste that accrues in areas maintained by municipalities
that are responsible for its disposal and recycling. People can throw gar-
bage in waste bins, which is why they are valuable in life [3]. If this did
not happen, the future would be a mess. Because a business or household
has a garbage disposal device, it becomes a valuable piece of equipment.
The dustbin’s position as a conciliator of changing waste practices has
barely been regarded, despite its importance in our daily lives. Bins, it is
believed, are providing a telling indicator of new garbage relationships in
society as they are repurposed as environmental technologies for mod-
ern recycling schemes. Garbage and litter are all over the television these
days, with disturbing statistics of debris filling the world. Despite the
grim news, a number of people and policymakers are trying to change
the trend by creative waste management practises. The Government has
encouraged city-based schemes and public-private collaboration projects
to improve waste management systems, but these have proven to be trou-
blesome. The shortage of financial capital, sufficient expertise, and techni-
cal competencies in the public sector are the main obstacles to developing
solid waste management systems around the world. Governments have
begun to look at PPPs as a possible solution. The amount of change and
development made was minimal. Medical waste management is tough
and complicated, especially during pandemics like COVID. Due to the
apparent forte of the global outbreak, contemporary waste centers are
adapting to reveal the abnormal scientific waste and its affect on viral
propagation with particular statistics on the amount of scientific waste
generated, waste warm spots, and remedy centers. Technological know-
how [24] on inspection, segregation, transportation, storage, and reliable
waste control structures are required to optimize contemporary sources
and centers to satisfy the crisis, as healthcare waste portions are antic-
ipated to increase rapidly. For patients, healthcare workers, and waste
Smart Health Care Waste Segregation and Safe Disposal 207
collectors, improper medical waste disposal can lead to accidents, dis-

eases, harmful consequences, and air pollution. Bio-dangerous cloth
and its opposite numbers encompass non-dangerous waste, infectious
waste, radioactive waste, bacterial waste, chemical waste, cytotoxic waste,
sharps waste, and pharmaceutical trash.
A pandemic has identified an unusual amount of medical waste. As a
result of this research, some serious problems were discovered and some
important suggestions were made. To prevent the spread of any dangerous
diseases, a proper waste management system is needed. This paper pro-
poses a waste management strategy that is both effective and successful.
This garbage sorter is made with sensors and an Arduino microcontroller.
The suggested scheme satisfies the need for continuous garbage material
tests in the bins. It aids in the disposal of garbage until the bins become
overflowing. As a result, the device is helpful in waste management when it
is monitored and informed on a daily basis. This translates to a cleaner city
and a higher quality of life.
14.2 Related Works

Garbage and litter are all over the television these days, with disturbing
statistics of debris filling the world. Despite the grim news, a number of
people and policymakers are trying to change the trend by creative waste
management practices. These five forward-thinking countries are taking a
novel approach to waste management in order to make the environment a
safer, healthier place. Germany is first, followed by Austria, South Korea,
Wales, and Indonesia. Clean Harbors, Stericycle Inc., Covanta Holding,
and others are among the best waste management firms in the world. The
Government of India has encouraged city-based schemes and public-
private collaboration projects to improve waste management systems,
but these have proven to be troublesome. The lack of financial resources,
appropriate skills, and technological competencies with the public sector
are the main obstacles to improving solid waste management services in
India. Governments have begun to look at PPPs as a possible solution.
The amount of change and development made was minimal. Some serious
problems have been discovered as a result of this research and some signif-
icant proposals have been made.
This article [4] presents a system (hardware, software, and communica-
tions) to improve trash handling while also involving citizens. The model
employs an IoT method in which the discharged trash from the trash
container is continuously monitored by sensors that provide real-time
information upon those filling levels of each chamber. An IoT-based smart

trash system [5] (SGS) is presented in this study to decrease food waste.
Wireless mesh networks enable battery-powered smart trash bins (SGBs)
to interact with one another in an SGS, while a router and server collect
and analyse data for application services [6], suggesting a unique approach
for achieving waste management that is both vigorous and efficient by fore-
casting the likelihood of waste levels in trash bins in this article. Combining
machine learning and graph theory, the system can optimise trash collec-
tion via the shortest path. This article describes an investigative case that
was carried out on the Ton Duc Thang University (Vietnam) campus to
assess the system’s performance and viability. A cross-domain robust dis-
tributed trust management [7] (Robust Trust) system is suggested in this
work that also makes a system suitable for independently evaluating faith
towards various devices. The credibility in this approach is split into three
security technologies that enable IoT nodes be resilient against hacked and
malicious devices/nodes. This study [8] proposes a strategy for achieving
this ambitious goal. In this article, a microcontroller is used to interface an
ultrasonic sensor and GSM modem to construct an intelligent container.
The highest point of the trashcan is fitted with an ultrasonic sensor, which
measures the dustbin’s height.
[9] A novel waste disposal based on an Android mobile app and a
Bluetooth-enabled smart dustbin is proposed in this article. Through the
lines painted on the floor, this android application controlled the bot. The
lines are usually specified along the path. A white line on a black surface or
vice versa might be one of the predetermined routes. In this study [10], a
novel architecture is suggested with the goal of improving waste disposal,
on-site handling, and transfer optimization. The network is based on a sen-
sor network and Data Transfer Nodes (DTN) are used to communicate
data from garbage bin filling to a remote server.
[4] This article presents a system (hardware, software, and communi-
cations) to improve trash management while also involving citizens. The
system uses an IoT method in which the discarded trash from the smart
bin is continually monitored by sensors that provide real-time information
on the filling level of each compartment [11]. This article proposes a revo-
lutionary sensor node design based on the use of low-cost, high-efficiency
components such as water level, soil moisture, temperature, humidity, and
rain sensors. The transmitting module, in particular, is based on the LoRa
LPWAN method, ensuring overall system performance. The principal cir-
cuit board of the system is optimised by combining two layers and doing
code optimization [12]. The concept of a garbage surveillance system for
smart campuses, colleges, clinics, and bus stops is proposed in this article.
The Garbage Monitoring System is a clever dustbin that makes use of sen-
sors to raise it above the trendy dustbin [13]. In the proposed system, pub-
lic trash cans would be equipped with an integrated device that allows for
real-time tracking of waste levels in the bins. The information on waste
levels would be used to create an optimal path for trash collection vans,
which will save money on fuel. The load sensors will enhance the accu-
racy of trash level information, while humidity sensors would give data on
waste segregation in a dust bin.
[14] The suggested model, which employs a convolutional neural net-
work (CNN), a prominent machine learning approach, to separate bio-
degradable and insoluble waste, is presented in this study. The plan also
includes an architectural concept for a smart garbage can that incorporates
a microprocessor and several sensors [15]. They presented an Internet of
Things-based management of solid waste in this article, which allows gar-
bage bin monitoring, dynamic scheduling, and tracking of garbage col-
lector vehicles in a city of the future. Garbage bins fitted with moderate
embedded devices are placed around the city under the proposed model.
[16] To combat COVID-19, this study suggests a leader-follower strat-
egy for hazardous infectious waste collection and government aid distribu-
tion. We suggest a smart waste material classification based on the 50-layer
residual net pre-train (ResNet-50) Convolutional Neural Network model,
which would be a machine learning tool that represents the extractor and a
Support Vector Machine (SVM) that is used to categorize waste into useful
groups/types including glass, metal, paper, and plastic among others. As
proposed by [17, 18], the “Smart Dustbin” in this article would be a cellu-
lar bin that autonomously monitors rubbish levels and transmits messages
to the appropriate city officials to update the bin’s status [19]. This article
discusses an automated system that allows depositors to dispose of their
e-waste and be compensated for it. In terms of e-waste collectors, an online
bidding session is held with the winner receiving ownership of the elec-
tronic trash that he or she won in the online bidding process. For this, an
interactive digital bin with multiple sensors and modules is required. After
user verification, the bin must be able to lock and unlock itself (Urlagunta).
The smart bin in this article is based on the Microcontroller system, which
is connected to a GSM modem and also an ultrasonic sensor. An ultra-
sonic sensor is mounted on the top of the trashcan to measure the dustbin’s
height. The minimum height is set at 10cm. The Microcontroller will be
designed such that when the trash fills, the remaining height above the
threshold height will be shown.
[20] This paper describes the design and implementation of an Internet
of Things (IoT) based Arduino microcontroller that uses ultrasonic
sensors to detect the amount of trash in garbage bins, reveal the infor-
mation, update it as “empty,” “half-filled,” or “filled” on an LCD screen
at periodic intervals, and also transmit the data level information [21].
An Internet of Things (IoT) architecture for real-time garbage monitor-
ing and collection is presented in this study, with the goal of improving
and optimizing solid waste collection in cities. The Netlogo multi-agent
platform was used to mimic real-time monitoring and intelligent waste
management decisions.
[22] An automatic trash segregator is proposed in this study. When
medical waste is detected, the conveyor belt is moved by an external motor.
The trash will be sent to the sensing and classification units. The input pic-
ture is captured, pre-processed, median filtered, contrast enhanced, and
segmented in five phases [23]. The article creates the ‘Waste Hierarchy-
Technology Readiness Levels’ framework and evaluates each waste man-
agement plan against it, showing the technological maturity and the
strategy’s place in the Waste Framework directive, as well as its relative
adherence to circular economy concepts [3]. This article presents a con-
cept for a smart garbage bin surveillance system in Ethiopian cities that
uses the Global System for Mobile Communication (GSM) to organize
trash collection networks in residential and commercial sectors. The level
of waste material in the trash bin has been monitored using an ultrasonic
sensor in this suggested system and it will continually connect with the
authorized clean-up person’s cell phone via GSM modem.
This research will be crucial not only for individuals whose professions
primarily entail the disposal of garbage cans, but also for the general pub-
lic who will be following the smart city rules for a higher quality of life.
The implementation of this suggested technique would go a long way for-
ward into ensuring effective and efficient real-time garbage disposal and
resource utilization, as well as improving life for smart city residents. The
actual implementation of the functional conceptual model will serve as an
advanced as well as instructive technique of sustaining and improving the
wellbeing of smart urban residents in long run.
14.3 System Architecture

Figure 14.1 illustrates the overall system architecture module. Inlet, Split
Rollers, Pipeline System, Dihydrogen Monoxide Outlet System, Exhaust
System, Chained Conveyor Belt, Incinerator, Solar Panel, Brushes,
Arduino Uno Microcontroller, IR Sensor, Ultrasonic Sensor, Moisture
Sensor, Glass Detector sensor, Inductive Proximity Sensor, Node-MCU
STEP DOWN
BRIDGE FILTER VOLTAGE
TRANSFORMER RECTIFIER CIRCUIT REGULATOR
(230 TO 12V AC) (IC 7805 & 7812)
POWER SUPPLY UNIT
DC MOTOR
GLASS (CONVEYOR
SENSOR BELT)
ULTRASONIC LEDS
SENSOR
MOISTURE ARDUINO
SENSOR ESP8266 MOBILE APP
PROXIMITY
SENSOR RELAY MOTOR
Figure 14.1 System architecture.
ESP8266, Wrapper, 3 chambers for dihydrogen monoxide, 2 for contam-

inant dihydrogen monoxide, and 1 for pristine dihydrogen monoxide,
Thermo-engenderer and Electrodes, ECU Board, Breadboards, Jumper
Wires, OLED Exhibit, LED Lights, and a DC Motor are a few of the main
components utilized for making the module.
If medical waste is disposed of in the inlet, it is transferred to the chain
conveyor belt, which triggers the DC motor, which causes the conveyor
belt to move. Metal, dry, wet, glass, and incinerated wastes are dissev-
ered into five categories. The sensors detect the waste predicated on the
type of waste and the waste is then sorted into concrete bins. The state
of filling the bin is immediately signaled by LEDs, utilizing artificial per-
spicacity. The red LED on the bin commences flashing as it approaches
full capacity and a caveat note is sent to the local ascendant entities. The
waste that has been filed is immediately bundled. The wastes that must
be incinerated are burned in the contrivance’s incinerator chamber. The
thermo-engenderer converts the heat engendered in the incinerator
chamber into electrical energy. The conveyor belt’s DC motor consumes
the engendered electrical energy. The conveyor belt is automatically
washed until the whole operation is consummated utilizing a chain con-
veyor belt cleaning machine.
IoT, AI, and web/app development are the three innovations utilized in
this scheme. To control the flow of waste into the conveyor, the inlet seg-
ment has an open and close function. To detect metallic waste, an induc-
tive proximity sensor is utilized. Arduino Uno is in charge of the conveyor
belt’s pacing and rotation. An incinerator is additionally part of this con-
trivance. When a human pushes the trigger, the controls are turned off
and the waste on the conveyor belt is sent to the incinerator container.
The walls of the incinerator are composed of clay and covered with alumi-
num foil. Inside the incinerator are the ECU board electrodes and thermo-
engenderer, which are habituated to engender electricity from thermal
energy. The DC motor is driven by the engendered electricity. The sensors
are deactivated until the procedure is done. This contrivance withal has
the capability of sending an admonishment SMS if a bin is loaded. Action
is detected by an infrared sensor. Moisture sensors for dry and wet waste,
Inductive Proximity sensors for metal waste, Ultrasonic sensors for bin
filling tracking, warning messaging via Node MCU ESP8266-12E, visual
exhibit via OLED exhibit, and waste packaging via hotwire sealer are all
utilized. The Biomedical Waste Segregator is a piece of automated machin-
ery that sorts waste into four categories: metal, glass, dry, and wet. The sug-
gested framework would be capable of monitoring and managing the solid
waste amassment process as well as the total amassment process. To detect
metallic waste, an inductive proximity sensor is utilized. Dry and wet waste
is disunited utilizing a blower system. A microcontroller controls the tim-
ing and rotation of the conveyor belt. This contrivance additionally has a
feature that sends an admonition SMS if a bin is loaded.
14.3.1 Wrapping
The Ultrasonic sensor causes the door of the respective container to
close as the waste is filled in it. The container’s door is made up of two
semi-circular panels. One of the circu lar plates is connected to the other
by a thin rod, while the other is connected by an insulating rod (namely
wood). A very low volt current is passing through each rod as the doors
close. The rods (the rod with electric flow heats up) come next to each
other until the door is fully locked. The heat and compression are just
enough to melt and seal the trash container. Once the doors of the con-
tainer are closed, the current in the rod flows till the rod reaches a high
enough temperature.
14.3.2 Incinerator
To burn the garbage, an incinerator is required. Medical waste must be
incinerated in some cases. When a human pushes the trigger, the con-
trols are turned off and the waste on the conveyor belt is sent to the
incinerator container. The walls of the incinerator are made of clay and
covered with aluminum foil. Inside the incinerator are the ECU board
electrodes and the thermos-generator, which are used to generate elec-
tricity from thermal energy. The DC motor is driven by the produced
electricity.
14.3.3 Conveyor Cleaning System

The conveyor cleaner makes contact with the returned conveyor and
uses a rotating, pre-moistened, heavy-duty belt to scrub it continuously.
For ease of product handling, the clean conveying surface is allowed
to dry. Conveyor cleaners are made of heavy-duty stainless steel and
have long-lasting scrubbing belts that need no maintenance. A 1/4
inch NTT-thread is used to connect to the water inlet and a manual
shutoff is included. A three-quarter inch ID hose is used to empty the
drain pan. The belt friction corrections, as well as the drain tub, are also
conveniently accessible for normal maintenance.The CC series of con-
veyor cleaners are enticing solutions for reduced tension line downtime
because of these characteristics, as well as ease of operation and easy
insulation.
14.3.4 Circuit Diagram

In Figure 14.2, the left most side ultrasonic sensor is connected trigg pin =
38 (output) and echo = 40 (input). On the left, the 2nd ultrasonic sensor
is connected trigg pin = 42 (output) and echo = 44 (input). The 3rd ultra-
sonic sensor is connected trigg pin = 46 (output) and echo = 48 (input).
On the right most side, the ultrasonic sensor is connected trigg pin = 52
(output) and echo = 50 (input). M2, M3, M4, and M5 are servo motors.
At M2, pwm=12. At M3, pwm=11. At M4, pwm=10. At M5, pwm=9. LED
2=51 LED 1 =49 LED 3 = 47 LED 4 =45. We use esp 8266. So, the tx is
connected to rx 0 and the rx is connected to tx 0, then the other pin of the
16 *2 LED display.
Figure 14.2 Circuit diagram.
14.3.5 Optimal Path Planning Algorithm for Waste

Collection
Figure 14.3 illustrates the flow chart of proposed methodology and the fol-
lowing algorithm supports the same.
Step 1: Set up the microcontroller and all of the sensors.

Step 2: Switch on the ESP8266 and initialise the SIM.
Step 3: When Wi-Fi is open, the mobile device connects to the
network using an IP address.
Step 4: When the height and weight of the bins exceed the mar-
gin, an SMS message is sent.
Step 5: Using the IP Address on the HTML tab, you can check
the status of the bins.
14.4 Methodology
The conveyor belt motor engages as the waste arrives and the conveyor
belt commences to move. Many of the motors and controls, as well as the
Start
Initializing all modules
Start taking data of sensors
Continuously monitor the level of the dustbin No

If know
status
No
If dry dust Yes Yes
bin/wet
dustbin is full Enter IP Address (192.168.4.1)
No Yes
Height
is full HTML page displayed
with details
No
If load is
full
Sending information to
authorized persons & server
Yes
(IoT)
Sending information to authorized
persons & server (IoT)
Figure 14.3 Optimal path planning algorithm.
microcontroller, have been switched on. With a significantly growing pop-

ulation and in this COVID pandemic, it is extensively more basic to be
benevolent concerning how well we, as individuals, manage our prosperity
and environment. Considering the insights, it is seen that authentic clini-
cal trash evacuation is especially expected for a spotless environment. The
modernized waste segregator is a capable and monetary waste combina-
tion structure with a base proportion of human mediation with no risk to
human life. Using a vehicle line makes the system significantly more accu-
rate, economic, and clearer to put in and use at a local level. Segregating
these misfortunes at a local level will be timesaving. The proposed struc-
ture fulfils the requirement for reliable monitoring of garbage content
in the containers. It helps with disposing of the waste material before it
floods from the canisters. So, standard observation and recommendation
make the structure significant in waste across the board. This prompts an
immaculate city for better living.
If the trash is not metallic, the conveyer belt sensor decides if it is a wet
or dry waste by estimating the dampness content of the waste. If there is
moisture in the waste, it is classified as wet waste and the conveyor belt is
switched off, the wet waste motor is turned on, and the garbage is depos-
ited in the wet waste container. Counter 2 is also increased. If the conveyor
belt is not carrying any wet trash, the waste is deposited into the dry waste
bin at the top of the conveyor belt. When a waste containing glass particles
is identified as glass waste, the ergo conveyor belt is turned off, the glass
waste motor is turned on, and the garbage is forced into the glass waste bin.
In consequence, Counter 3 is raised. The Figure 14.4 brings about the flow
diagram of overall process.
Conclusively, the wastes are deposited in the congruous containers,
consummating the segregation process. When the dustbin level reaches
50%, the yellow LED which is annexed to the bins gets turned on. At 75%,
the green LED gets turned on. At 90%, the red LED commences blink-
ing and at 100%, the red LED gets turned on and then the SMS alert will
be activated. When the bin gets filled, by utilizing the wrapper, the waste
gets wrapped automatically. This system has a self-cleaning system with
a conveyor belt. The conveyor chain enters the conveyor cleaner and is
Medical waste.
Split roller splits the Medical waste.
Waste is passed onto the conveyer belt.
The sensors detect the waste and send into the particular bin.
Once the bin is filled, waste get warped.
Self cleaning of the conveyer belt will be done automatically.
The thermal energy which comes from the increnator will produce the electricity.
When the bin is filled GMC get intimated automatically.
It produces high accuracy in segregating the waste.
Eco-friendly & hygenic.
Figure 14.4 Flow diagram of overall process.

engaged by the hold-down bracket. The hold-down bracket guarantees that

the chain makes full contact with the scrubbing belt, which rotates in the
antithesis direction. The conveyor chain is squeegeed by a spring-loaded
neoprene blade as it exits the conveyor cleaner before being returned to the
conveyor line. The conveyor belt is swept in this manner.
14.5 Mobile App
Figure 14.5 Web page inerface to display status of smart bins.

This Mobile App as seen in Figure 14.5 describes about detecting the
amount of waste in each bin. First, we need to select the state, district, and
area. Using this information, it displays the number of bins present in that
area and the amont of waste in them. As mentioned, LEDs are used for easy
identification of the amount of waste. If it is of red color, it indicates that
the bin is completely filled and needs to be cleaned. If it is yellow the bin is
partially filled and if it is green it is empty. This information is sent to the
nearest municipal corporations, so that the waste disposal becomes easy.
14.6 Conclusions and Future Works

With a dramatically expanding populace and in this Coronavirus pan-
demic, it is considerably more critical to be kind with regards to how well
we, the people, deal with our wellbeing and climate. In view of the percep-
tions, it is perceived that legitimate clinical garbage removal is particularly
required for a clean climate. The computerized squander segregator is a
proficient and financial waste assortment framework with a base measure
of human intercession and furthermore makes no peril for human exis-
tence. Utilizing a transport line makes the framework substantially more
exact, financially savvy, and furthermore more straightforward to place in
and use at a homegrown level. Isolating these losses at a homegrown level,
likewise, will be timesaving. The proposed framework satisfies the need for
consistently keeping an eye on trash content in the receptacles. It assists
with discarding the waste material before it floods from the canisters. So,
standard observation suggests that the framework is valuable across the
board. This prompts a spotless city for better living.
The automatic waste segregator is a cost-efficacious and reliable waste
accumulation contrivance that requires no human involution and poses
little risk to human safety. The utilization of a conveyor belt ameliorates
the precision, cost-efficacy, and facilitation of installation and utilization
of the system on a domestic substructure. Dissevering these wastes at the
household level would preserve time. The suggested scheme is slated for
perpetual monitoring of the medical waste material in the bins. It avails in
the disposal of garbage until the bins become overflowing. This translates
to a cleaner city and a higher standard of living.
Restorative treatment is basic to our survival and supportability.
Incomprehensible restorative squander isolation at the point of inchoa-
tion may have a domino impact on the environment, posturing dangers
to people, natural life, and soil, as well as dihydrogen monoxide bodies.
Natural dangers related with destitute healthcare squander administration
may sully the air we breathe through poisonous airborne contaminants in

the event that they are not satisfactorily contained, isolated, and burned by
on-site or off-site burning. As a result, such squander needs uncommon
care and support before being disposed of. It is vital that healthcare labour-
ers get it the esteem of therapeutic squander control.
Declarations
Availability of Data and Materials
Information accessibility is not appropriate for this composition, as no
unused information was made or analyzed in this ponder.
Competing Interests
The Author(s) declare(s) that there is no conflict of interest.
Author’s Contribution
All the authors contributed indistinguishably to the work. All authors
examined and endorsed the ultimate composition.
References
1. B. Bharadwaj, M. Kumudha, N. Gowri Chandra, G. Chaithra, Automation of
Smart waste management using IoT to support “Swachh Bharat Abhiyan” -
A practical approach, Proc. 2017 2nd Int. Conf. Comput. Commun. Technol.
ICCCT 2017. (2017) 318–320. https://doi.org/10.1109/ICCCT2.2017.7972300.
2. A. Sharma, R. Kumar, V. Mansotra, Proposed Stemming Algorithm for
Hindi Information Retrieval, Int. J. Innov. Res. Comput. Commun. Eng.
(An ISO Certif. Organ. 3297 (2016) 11449–11455. https://doi.org/10.15680/
IJIRCCE.2016.
3. M. Karthikeyan, D. Eligo, S. Gebrehiyot, E. Bekele, D. Daniel, G. Dansa, GSM
Based Smart Waste Bin Monitoring System In Ethiopia, VI (2019) 796–804.
4. K. Pardini, J.J.P.C. Rodrigues, O. Diallo, A.K. Das, V.H.C. de Albuquerque,
S.A. Kozlov, A smart waste management solution geared towards citizens,
Sensors (Switzerland). 20 (2020) 1–15. https://doi.org/10.3390/s20082380.
5. I. Hong, S. Park, B. Lee, J. Lee, D. Jeong, S. Park, IoT-Based Smart Garbage
System for Efficient Food Waste Management, Sci. World J. 2014 (2014).
https://doi.org/10.1155/2014/646953.
6. T. Anh Khoa, C.H. Phuc, P.D. Lam, L.M.B. Nhu, N.M. Trong, N.T.H. Phuong,
N. Van Dung, N. Tan-Y, H.N. Nguyen, D.N.M. Duc, Waste Management
System Using IoT-Based Machine Learning in University, Wirel. Commun.
Mob. Comput. 2020 (2020). https://doi.org/10.1155/2020/6138637.
7. K.A. Awan, I. Ud Din, A. Almogren, M. Guizani, A. Altameem, S.U. Jadoon,
RobustTrust - A Pro-Privacy Robust Distributed Trust Management
Mechanism for Internet of Things, IEEE Access. 7 (2019) 62095–62106.
https://doi.org/10.1109/ACCESS.2019.2916340.
8. V.P. Vijaynaidu, T. Dhikhi, Smart garbage management system, Int. J. Pharm.
Technol. 8 (2016) 21204–21211. https://doi.org/10.17577/ijertv4is031175.
9. C.S. Anilkumar, G. Suhas, S. Sushma, A smart dustbin using mobile appli-
cation, Int. J. Innov. Technol. Explor. Eng. 8 (2019) 3964–3967. https://doi.
org/10.35940/ijitee.K2129.0981119.
10. S. Longhi, D. Marzioni, E. Alidori, G. Di Buò, M. Prist, M. Grisostomi, M.
Pirro, Solid waste management architecture using wireless sensor network
technology, 2012 5th Int. Conf. New Technol. Mobil. Secur. - Proc. NTMS
2012 Conf. Work. (2012). https://doi.org/10.1109/NTMS.2012.6208764.
11. T.A. Khoa, M.M. Man, T.Y. Nguyen, V.D. Nguyen, N.H. Nam, Smart agricul-
ture using IoT multi-sensors: A novel watering management system, J. Sens.
Actuator Networks. 8 (2019). https://doi.org/10.3390/jsan8030045.
12. R. Asha, N. Mahajan, A. Yadav, K.S. Balamurugan, Garbage Managing Smart
System using-IOT, Int. J. Recent Technol. Eng. 8 (2019) 3646–3649. https://
doi.org/10.35940/ijrte.d7849.118419.
13. P.S.A. Mahajan, A. Kokane, A. Shewale, M. Shinde, S. Ingale, Smart Waste
Management System using IoT, Int. J. Adv. Eng. Res. Sci. 4 (2017) 93–95.
https://doi.org/10.22161/ijaers.4.4.12.
14. M.W. Rahman, R. Islam, A. Hasan, N.I. Bithi, M.M. Hasan, M.M. Rahman,
Intelligent waste management system using deep learning with IoT, J.
King Saud Univ. - Comput. Inf. Sci. (2020). https://doi.org/10.1016/j.
jksuci.2020.08.016.
15. S.S. Chaudhari, V.Y. Bhole, Solid Waste Collection as a Service using IoT-
Solution for Smart Cities, 2018 Int. Conf. Smart City Emerg. Technol.
ICSCET 2018. (2018) 1–5. https://doi.org/10.1109/ICSCET.2018.8537326.
16. J. Valizadeh, A. Hafezalkotob, S.M. Seyed Alizadeh, P. Mozafari, Hazardous
infectious waste collection and government aid distribution during COVID-
19: A robust mathematical leader-follower model approach, Sustain. Cities
Soc. 69 (2021) 102814. https://doi.org/10.1016/j.scs.2021.102814.
17. O. Adedeji, Z. Wang, Intelligent waste classification system using deep learn-
ing convolutional neural network, Procedia Manuf. 35 (2019) 607–612.
https://doi.org/10.1016/j.promfg.2019.05.086.
18. Aniqa Bano, Ikram Ud Din, Asma A. Al-Huqail, “AIoT-Based Smart Bin
for Real-Time Monitoring and Management of Solid Waste”, Scientific
Programming, vol. 2020, Article ID 6613263, 13 pages, 2020. https://doi.
org/10.1155/2020/6613263.
19. K.B. Kumar R, S.B. Patil, S. Manjula, H.N. Kumar, An IoT Based Approach
for Efficient Collection and Disposal of E-waste, Int. J. Eng. Tech. 4 (2018)
379–383.
20. S. Abba, C. Light, IoT-Based Framework For Smart Waste Monitoring And
Control System, A Case Study of Smart Cities., (2020) 8224. https://doi.
org/10.3390/ecsa-7-08224.
21. E.D. Likotiko, D. Nyambo, J. Mwangoka, M Ulti -a Gent B Ased I Ot S Mart
W Aste M Onitoring a Nd C Ollection, International Journal of Computer
Science, Engineering and Information Technology (IJCSEIT), Vol.7, No.5,
October 2017.
22. K.G. Devi, K. Yasoda, M. Dhivya, B. Kishore, Automatic Health Care Waste
Segregation and Disposal System, J. Xidian Univ. 14 (2020) 5281–5290.
https://doi.org/10.37896/jxu14.5/573.
23. C.A. Fletcher, R. St. Clair, M. Sharmina, A framework for assessing the cir-
cularity and technological maturity of plastic waste management strategies
in hospitals, J. Clean. Prod. 306 (2021) 127169. https://doi.org/10.1016/j.
jclepro.2021.127169.
24. Sarkodie, S.A., Owusu, P.A. Global assessment of environment, health
and economic impact of the novel coronavirus (COVID-19). Environ Dev
Sustain 23, 5005–5015 (2021). https://doi.org/10.1007/s10668-020-00801-2.
15
Investigation of Viscoelastic
Magnetohydrodynamics (MHD)
Flow Over an Expanded Lamina
Surrounded in a Permeable Media
Hiranmoy Mondal1*, Arindam Sarkar2 and Raj Nandkeolyar2
1
Department of Applied Science, Maulana Abul Kalam Azad University of
Technology, West Bengal, Kolkata, India
2
Department of Mathematics, National Institute of Technology Jamshedpur,
Jamshedpur, India
Abstract
The results of the boundary value problem of MHD nanofluid micropolar fluid
flow have been carried out throughout this study. Using the spectral quasi-
linearization method (SQLM), the mass as well as heat transfer effects of fluid have
been investigated. A stream consists of viscous dissipation, bouncy force, thermal
radiation, and Joule heating. It has been scrutinized that the existence of magnetic
parameter Mn enhances concentration and temperature gradient, whereas the
appearance of permeable media turns down the velocity profile.
Keywords: Stretching surface, MHD flow, darcy dissipation, viscoelastic liquid,

nanofluids, chemical reaction, spectral quasi linearization method
15.1 Introduction
15.1.1 Literature Review
Viscoelastic Magneto-hydrodynamic nanofluids through a stretching
sheet have a broad range of practical applications in industries such as
paper production, metal spinning, relegation of plastic lamina, continuous
*Corresponding author: hiranmoymondal@yahoo.co.in
223
welding of metals, production of synthetic sheets, removal of heat from

metallic sheets, etc.
The study of magnetohydrodynamics (MHD) takes place as an essential
part in thermodynamics, especially during the determination of boundary
layer flow for cooling of nuclear reactors, high-temperature plasmas, etc. [1].
Pal and Mondal [2] reported that the stretching parameter and velocity pro-
file are inversely related. Baag et al. [3] investigated the entropy generation
through an electrically-conducting viscoelastic MHD fluid stream upon an
exponentially expanded porous lamina. The boundary layer equations have
been solved via Kummer’s function. The viscoelastic and magnetic parame-
ters were shown to be inversely related to velocity, whereas the elastic param-
eter was linearly proportional to velocity. Furthermore, porous matrix and
the inverse of the Prandtl number were found to be directly proportional
to temperature [3]. Also, Pal and Mondal [4] studied laminar, unwavering,
and convective boundary layer Newtonian fluid streams. They examined the
heat absorbing/producing effect of the fluid which is passing through a non-
isothermal along with the thermal radiation immersed in a permeable surface.
15.1.2 Nomenclature
B0 Strength of uniform magnetic field
Cp Specific heat of the fluid
f Stream function
K Thermal conductivity of the fluid
Mn Magnetic parameter
R Radiation parameter
T Non-dimensional temperature
T∞ Ambient temperature/temperature far from sheet
ρ Density of the fluid
Kp Porosity parameter
k0 Dimensionless elastic parameter
K ′p Porosity parameter
s Plate concentration parameter
r Plate temperature parameter
D Molecular diffusivity
DB Brownian motion coefficient
DT Thermophoretic diffusion coefficient
µ Dynamic viscosity
ν Kinematic coefficient of viscosity
σ Electrical conductivity
Rc Elastic parameter
Pr Prandtl number
Viscoelastic MHD Flow Over an Expanded Lamina 225
Nt Thermophoresis parameter
Nb Brownian motion parameter
Le Lewis number
At present, nanofluids are very significant in engineering applications

as well as biomedical science, as it increases heat transfer and energy effi-
ciency in various kinds of thermal systems. Nanofluid is a special kind of
heat transfer medium containing oxides or metals, e.g., zinc oxide, copper
oxide, alumina or iron oxide having a radius less than 100 nm, which are
equally and stably spread out in a base fluid like water, oil, acetone, etc.
Over the last few years, nanofluids have drawn a substantial amount of
attention for their unique properties which are very beneficial in the chem-
ical and metallurgical area, cooling, ventilation, thermal therapy for cancer
treatment, power generation, micro-manufacturing, and air conditioning.
The literature survey shows that during the past few decades, curiosity
around fluid flowing through a stretching surface has grown. Crane [5]
investigated constant stretching surface temperature problems. Later on, a
small number of researchers investigated the outcomes of chemical reper-
cussions, heat and mass alternation, thermal slip conditions of various
non-Newtonian fluids, or probable combined effects of the above [6–11].
Due to countless industrial applications like power generation system
and modern metallurgical processes, many researches have been attracted
to the MHD electrically conducting fluid flow. Hayat et al. [12] considered
an unperturbed incompressible MHD Walter-B nanofluid flow that is pass-
ing through a nonlinear stretching surface. Using homotopy analysis they
deciphered their mathematical model, whereas Daniel et al. [13] did his
research in the heat alternation effect of MHD nanofluid stream take across
a porous stretching surface. Hayat et al. [14] studied the heat transmission
of a nanofluid stream with various dissipation and Joule heating along a
permeable stretched cylinder and the stream was subjected to nonlinear
thermal diffusion with homogeneous-heterogeneous reactions. They used
an Explicit Euler formula to resolve their fluid model. Hayat et al. [15] cal-
culated the effect of melting heat distribution of a MHD nanofluid genre
with nonlinear thermal diffusion in regards to the stagnation point along
an expanded lamina. Using homotopy analysis they deciphered their flow
equations. Kishan and Maripala [16] presented a mathematical model of the
Darcy Forchheimer MHD boundary layer stream by a permeable expanded
lamina. They also calculated the thermophoresis and viscous dissipation.
The perusal of a hydromagnetic stream of an electrically conducting fluid
has drawn a significant observation due to its applications in recent met-
allurgical and metal-working processes. The effect of a magnetic field is an
important part to manage the momentum as well as heat conduction in the
boundary layer stream of various kind of fluids. Several researchers (Ibrahim

et al. [17], Turkyilmazoglu [18], Farooq et al. [19], Sheikholeslami et al. [20],
Baag et al. [21], Sheikholeslami et al. [22], Abdul Hakeem et al. [23]) studied
about the impact of magnetic field on MHD fluid.
It is immensely important to scrutinize the behavior of heat absorp-
tion/production on account of several kinds of physical models in mod-
ern days. Eldahab and Aziz [24] have investigated the results of uneven
heat exploitation/whiff effect of viscous fluid with thermal radiation. Abel
et al. [25] examined the non-Newtonian MHD viscoelastic fluid upon a flat
lamina. Ganga et al. [26] investigated the entropy generation on nanofluid
flow through an expanded lamina by considering velocity slip condition
and nonlinear thermal radiation. The study of heat distribution analysis for
multiphase magnetic fluid passing through a flat lamina with heat exploita-
tion/production and thermal diffusion was scrutinized by Zeeshan et al.
[27]. The spectral quasi-linearization method (SQLM) is an important
numerical method to solve highly nonlinear differential equation in fluid
dynamics. Mondal et al. [28] and Motsa [29] used this method to solved
the mathematical problem of the fluid flow.
The originality of the recent perusal is to consider the chemical reaction and
thermophoresis effect on the viscoelastic MHD nanofluid flow. The values of
(Cf), (Nu), and (Sh) for different parameters has been discussed numerically.
The outcome of this study will provide useful information for applications in
several engineering processes along with a complement to the prior perusal.
15.2 Formulation of the Problem

Along the way of the main stream of the lamina, the x − axis has been
considered and the direction of the y – axis is perpendicular to the lamina
with velocity components u, v along the x – axis and y – axis, respectively
as demonstrated in Figure 15.1.
The ruling equations of two-dimensional stream considering the bound-
ary layer approximation are presented in this fashion:
∂u ∂v
+ =0 (15.1)
∂x ∂y
∂u ∂u ∂ 2u B 2u v
u +v = v 2 −σ 0 − u
∂x ∂r ∂y ρ K P′
k0  ∂  ∂ 2u  ∂ 3u ∂u ∂ 2u 
−   u  + v −  (15.2)
ρ  ∂x  ∂y 2  ∂y 3 ∂y ∂x∂y 
Vetical Magnetic field
B0 Momentum Boundary Layer
T∞ C∞
Thermal Boundary Layer
y
Porous Medium
Concentration Boundary layer
Slit Stretching sheet up = λx
Tw Cw
Figure 15.1 Schematic diagram.
 ∂T ∂T  ∂ 2T  ∂T ∂C DT  ∂T 2 
ρC p  u +v = k 2 + τ  DB +  (15.3)
 ∂x ∂y  ∂y  ∂y ∂y T∞  ∂y  
∂C ∂C ∂ 2C D ∂ 2T (15.4)
u +v = D 2 + T 2 − R(C − C∞ )
∂x ∂y ∂y T∞ ∂y
The relevant boundary conditions are

r
x
=
u u=
p λ x, = , T Tp (x ) = A   + T∞ ,
v 0=
l
s
x
=C C= p (x ) B   + C∞ , at y = 0 (15.5)
l
∂u
u = 0, = 0, T → T∞ , C → C∞ as y → ∞.
∂x
15.2.1 Analytical Solution

The stream function Ψ(x, y) is worthwhile for the continuity equation
∂Ψ ∂Ψ
u= ,v = − . (15.6)
∂x ∂y
Considering the similarity conversion
λ
=η y= , Ψ (x , y ) x vλ f (η )
ν
Substituting all the above values, Equations (15.2), (15.3), and (15.4) are
transformed as follows:
 1 
f ′′′ + ff ′′ − f ′2 − Rc{2 f ′ f ′′′ − ff ′′2 − ff iv } −  Mn +  f ′ = 0 (15.7)
 KP 
θ″ + {Nbθ′∅′ + Ntθ′2 – (rf′θ − fθ′)}Pr = 0 (15.8)
1 1 Nt (15.9)
sf ′∅ − f ∅′ = ∅′′ + θ ′′ − R1∅
Le Le Lb
f(0) = 0, f′(0) = 1, f′(∞) = 0, f″(∞) = 0,
θ(0) = 1, θ(∞) = 0
∅′(0) = −1, ∅ → 0 at η = ∞ (15.10)
Wherein f′(η) stand for the differentiation of f with respect to η,

τ DB ∆C τ D ∆T
Nb = represents Brownian motion parameter, N t = T repre
v v vT∞
sents the thermophoresis parameter, Pr = is the Prandtl number,
v R k
Le = is the Lewis number, and R1 = .
D λ
15.2.2 Numerical Methods (Spectral Quasi-Linearization

Methods)
Amalgamated with boundary condition (15.10), the nonlinear-coupled
ordinary differential Equations (15.7), (15.8), and (15.9) are solved numer-
ically, applying the spectral quasi-linearization method (SQLM).
 1 
F ≡ f ′′′ + ff ′′ − f ′2 − Rc{2 f ′f ′′′ − f ′′2 − ff iv } −  Mn +  f′=0
 KP 
∂F
a0, r = iv = Rcf r
∂f
∂F
a1, r = = 1 − 2Rcf r′
∂f ′′′
∂F
a2, r = = f r + 2Rcf r′′
∂f ′′
∂F  1 
a3, r = = −2 f r′ − 2Rcf r′′′−  Mn +
∂f ′  K P 
∂F
a4 , r = = f r′′+ Rcf riv
∂f
The linear term is as follows:
a0, r f riv+1 + a1, r f r′′′+1 + a2, r f r′′+1 + a3, r f r′+1 + a4, r f r =

RF
The nonlinear term of the energy equation is:
θ ≡ θ ′′ + {N bθ ′∅′ + N tθ ′2 − (rf ′θ − f θ ′)}Pr = 0

∂θ
b0 , r = =1
∂θ ′′
∂θ
b1, r = = (N b∅′r + 2N tθr′ + f r )Pr
∂θ ′
∂θ
b2 , r = = −Pr rf ′
∂θ
∂θ
b 3, r = = −Pr rθr
∂f ′
∂θ
b4, r = = Prθ ′
∂f
∂θ
b5, r = = N b Prθ ′
∂∅′
The converted linear term is as follows:
b0, rθr′′+1 + b1, rθr′+1 + b2, rθr +1 + b3, r f r′+1 + b 4, r f r +1 + b5, r ∅′r +1 = Rθ
The nonlinear term of the concentration equation is:
Nt
∅ ≡ ∅′′ + θ ′′ − LeR1∅ − Le(sf ′∅ − f ∅′)
Nb
∂∅
c0, r = =1
∂∅′′
∂∅
c1, r = = Lef r
∂∅′
∂∅
c2, r = = −LeR1 − sLe f r′
∂∅
∂∅
c 3, r = = −sLeθr
∂f ′
∂∅
c 4, r = = Le∅′r
∂f
∂∅ N b
c 5, r = =
∂θ ′′ N t
The converted linear term is as follows:
c0, r ∅′′r +1 + c1, r ∅′r +1 + c2, r ∅r +1 + c3, r f r′+1 + c 4, r f r +1 + c5, rθr′′+1 = R∅
15.3 Result and Argument

The effect of ruling parameters on the skin-friction coefficient, Nusselt
number, and Sherwood number has been presented in the following, Table
15.1 from tabular data we observe that (Cf) enhances with a rise in both
magnetic parameter and thermophoresis parameter, whereas (Cf) reduces
with enhancement in Soret number, chemical repercussion, and Brownian
motion parameter. Nu reduces monotonically with enhancement in
Soret number and rises with enhancement in magnetic, Brownian, and
Table 15.1 Values (Cf), (Nu), and (Sh) for different parameters.
Mn Sr R1 Nb Nt −f″(0) −θ′(0) −Ø′(0)
0.0 0.7 0.2 0.5 0.2 6.47747836 -1.24792674 -0.16130247
0.2 6.68965205 -1.23881769 -0.15872752
0.6 7.09600494 -1.22067600 -0.15482334
0.2 0.7 0.2 0.5 0.2 6.68965205 -1.23881769 -0.15872752
1.6 6.68959350 -2.06833985 0.15987681
5.0 6.69009405 -4.33945643 1.04495979
0.2 0.7 0.0 0.5 0.2 6.68959327 -1.57062405 0.21775166
0.4 6.69037312 -1.09209497 -0.36582335
0.6 6.69002619 -1.00077201 -0.51685468
0.2 0.7 0.2 0.2 0.2 6.68988106 -1.67046286 0.80664545
0.6 6.68993581 -1.12522670 -0.25586930
0.8 6.69012545 -0.93800745 -0.36733042
0.2 0.7 0.2 0.5 0.1 6.68959286 -1.32703668 -0.32973517
0.4 6.68990575 -1.09355804 0.11126215
0.7 6.69039059 -0.93732032 0.40484395
thermophoresis parameters and chemical reaction. Further, it has been

audited that the sequel of magnetic parameter, thermophoresis parameter,
and Soret number is to enhance the Sherwood number and reverse the
effects of chemical reactions. The Brownian motion parameter has been
observed on the Sherwood number.
Figures 15.2 and 15.3 show a graph of Mn on f′(η) and f(η), respectively.
Through Figure 15.4 it can be audited that the result of Mn enhances the
concentration at all points.
Figures 15.5 and 15.6 portray the temperature gradient θ(η) for several
values of the magnetic parameter Mn and Prandtl number Pr. It has been
audited that in Figure 15.5, θ(η) increases with η and it does not depend
on the value of the Prandtl number, thermophoresis parameter, Lewis
parameter, etc., whereas Figure 15.7 portrays an enhancement in the elas-
tic parameter Rc, which decreases the transverse velocity f(η) for Mn = 2.
0.8
Rc = 0.1; Kp = 0.5; Sr = 0.2; S = 0.5;
Pr = 8; Nb = 0.5; Nt = 0.2;
Le = 0.6; R1 = 0.5;
0.6
Mn = 0.1
f’(η)
Mn = 2.5
Mn = 3.5
0.4
Mn = 4.7
0.2
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
η
Figure 15.2 Result of Mn on f′(η) for Rc = 0.1, Pr = 8.
0.6
0.5
0.4
0.3
f(η)
Rc = 0.1; Kp = 0.5; Sr = 0.2; S = 0.5

Pr = 8; Nb = 0.5; Nt = 0.2; Le = 0.6; Che = 0.5
0.2
Mn = 0.1
0.1 Mn = 2.5
Mn = 3.5
0 Mn = 4.7
-0.1
0 1 2 3 4 5 6 7 8 9 10
η
Figure 15.3 Results of Mn on f(η) for Rc = 0.1, Pr = 8.
0.8 Rc = 0.1; Kp = 0.5; Sr = 0.2; S = 0.5;

Pr = 8; Nb = 0.5; Nt = 0.2; Le = 0.6; Che = 0.5
0.6 Mn = 0.1
Mn = 2.5
Φ(η)
0.4 Mn = 3.5
Mn = 4.7
0.2
-0.2
0 1 2 3 4 5 6 7 8 9 10
η
Figure 15.4 Results of Mn on Ø(η) for Rc = 0.1, Pr = 8.

0.8
Rc = 0.1; Kp = 0.5; Sr = 0.2; S = 0.5;
Pr = 8; Nb = 0.5; Nt = 0.2; Le = 0.6; Che = 0.5
0.6 Mn = 0.1
θ(η)
Mn = 2.5
Mn = 3.5
0.4 Mn = 4.7
0.2
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
η
Figure 15.5 Results of Mn on θ(η) for Rc = 0.1, Pr = 8.
0.8
Mn = 2; Kp = 0.5; Sr = 0.2; S = 0.5
Rc = 0.1; Nb = 0.5; Nt = 0.2; Le = 0.6; Che = 0.5
0.6
θ(η)
Pr = 6.8
Pr = 8.9
0.4 Pr = 10
Pr = 12
0.2
0
0 1 2 3 4 5 6 7 8 9 10
η
Figure 15.6 Results of Pr on θ(η) for Mn = 2.
0.5
0.4
Mn = 2; Kp = 0.5; Sr = 0.2; S = 0.5;
0.3 Pr = 8; Nb = 0.5; Nt = 0.2; Le = 0.6; Che = 0.5;
Rc = 0
f(η)
0.2 Rc = 0.01
Rc = 0.1
0.1 Rc = 0.2
-0.1
0 1 2 3 4 5 6 7 8 9 10
η
Figure 15.7 Results of Rc on f(η) for Mn = 2.

The concentration profiles Ø(η) for various values of chemical reper-

cussion parameter Che are portrayed in Figure 15.8. The temperature and
chemical reaction parameters are inversely related.
The influence of the plate concentration parameter S, Lewis parameter
Le, and Porosity parameter Kp on the concentration profile is shown in
Figure 15.9, 15.10, and 15.11. From Figures 15.8 and 15.9, it is audited that
concentration gradient parameter Ø(η) diminishes with η for Mn keep-
ing constant, Mn = 2, i.e., the plate concentration parameter and Lewis
parameter are inversely proportional with concentration gradient. It is also
audited that with the enhancement on η, after a certain time there will
be no effect of plate concentration parameters on concentration gradient.
Figure 15.11 illustrates the result of porosity parameter on velocity gradi-
ent and it is seen that appearance of porous matrix increases the velocity
0.8
Mn = 2; Kp = 0.5; Sr = 0.2; S = 0.5;
Pr = 8; Nb = 0.5; Nt = 0.2; Le = 0.6; Rc = 0.1
0.6
Φ(η)
Che = 0.5
Che = 2
0.4 Che = 5
Che = 8
0.2
0
0 1 2 3 4 5 6 7 8 9 10
η
Figure 15.8 Results of chemical reaction on Ø(η) for Mn = 2, Rc = 0.1, S = 0.5, Pr = 8.
0.8
Mn = 2; Kp = 0.5; Sr = 0.2
0.6 Pr = 8; Nb = 0.5; Nt = 0.2;
Le = 0.6; Rc = 0.1; Che 0.5
S = 0.5
Φ(η)
0.4 S=2
S=8
0.2 S = 10
-0.2
0 1 2 3 4 5 6 7 8 9 10
η
Figure 15.9 Results of plate concentration on Ø(η) for Mn = 2, Rc = 0.1, Che = 0.5.
0.8
Mn = 2; Kp = 0.5; Sr = 0.2;
Pr = 8; Nb = 0.5; Nt = 0.2;
0.6 Le = 0.6; Rc = 0.1; Che = 0.5, S = 0.5
Le = 0.6
Φ(η)
0.4 Le = 1
Le = 2
Le = 5
0.2
-0.2
0 1 2 3 4 5 6 7 8 9 10
η
Figure 15.10 Results of Lewis parameter Le on Ø(η) for Mn = 2, Rc = 0.1, Che = 0.5.
0.6
0.5
0.4
0.3 Mn = 2; Sr = 0.2; Pr = 8; Nb = 0.5; Nt = 0.2;
f(η)
Le = 0.6; Rc = 0.1; Che = 0.5, S = 0.5
0.2
Kp = 0.5
Kp = 1
0.1
Kp = 1.6
Kp = 2.5
0
-0.1
0 1 2 3 4 5 6 7 8 9 10
η
Figure 15.11 Results of Kp on f(η) for Mn = 2, Rc = 0.1, Che = 0.5.
at each point and after a certain time, transverse velocity becomes parallel
to the axis of η.
15.4 Conclusion
The solutions for a steady boundary layer stream and heat distribution for
a nanofluid throughout an exponentially expanded lamina in the presence
of a chemical reaction is analysed. Results of the magnetic parameter and
the viscoelastic parameter on various velocities are discussed. The impres-
sion of the magnetic parameter, Prandtl number, and the heat sink/source
parameter on the temperature profiles are presented. The main findings of
this research may be shortened as follows:
a) Concentration and temperature are directly proportional to

the magnetic parameter
b) Elastic parameter decreases the temperature at every point
c) Plate concentration parameter and Lewis parameter are
inversely proportional with concentration profile
d) Presence of magnetic parameter and porous matrix decreases
the velocity at all points
References
1. D. Pal and H. Mondal, Influence of thermophoresis and Soret-Dufour on
magnetohydrodynamic heat and mass transfer over a non-isothermal wedge
with thermal radiation and Ohmic dissipation, J. Magn. Magn. Mater., 331,
250–255, 2013.
2. D. Pal and H. Mondal, Influence of Soret and Dufour on MHD buoyancy-
driven heat and mass transfer over a stretching sheet in porous media with
temperature-dependent viscosity, Nucl. Eng. Des., 256, 350–357, 1970.
3. S. Baag, S. R. Mishra, G. C. Dash, and M. R. Acharya, Entropy generation
analysis for viscoelastic MHD flow over a stretching sheet embedded in a
porous medium, Ain Shams Eng. J., 8, no. 4, 623–632, 2017.
4. D. Pal and H. Mondal, Influence of temperature-dependent viscosity and
thermal radiation on MHD forced convection over a non-isothermal wedge,
Appl. Math. Comput., 212, no. 1, 194–208, 2009.
5. Crane LJ. Flow past a stretching plane. Jape Math Phys (ZAMP), 21, 645–7,
1970.
6. Ishak A, Nazar R, Pop I. Heat transfer over a stretching surface with variable
heat flux in micropolar fluids. Phys Lett A, 372, 559–61, 2008.
7. Ishak A, Nazar R, Arifin NM, Pop I. Mixed convection of the stagnation
point flow towards a stretching vertical permeable sheet. Malaysian J Math
Sci 2, 217–226, 2007.
8. Ishak A, Nazar R, Pop I. Mixed convection boundary layers in the stagna-
tion point flow towards a stretching vertical permeable sheet. Mechanica 41,
509–518, 2006.
9. Yao S, Fang T, Zhong Y. Heat transfer of a generalized stretching/shrinking
wall problem with convection boundary conditions. Commun Nonlinear Sci
Numer Simul 16, 752–760, 2011.
10. Quasim M, Hayat T, Obaidat S. Radiation effect on the mixed convection
flow of a viscoelastic fluid along an inclined stretching sheet. Z Naturforsch
67, 195–202, 2012.
11. Hayat T, Quasim M, Mesloub S. MHD flow and heat transfer over a stretch-
ing sheet with Newtonian heating with slip condition. Int J Numer Methods
Fluid 66, 963–975, 2011.
12. Hayat, T., Qayyum, S., Alsaedi, A. and Ahmad, B., Magnetohydrodynamic
(MHD) nonlinear convective flow of Walters-B nanofluid over a nonlinear
stretching sheet with variable thickness, International Journal of Heat and
Mass Transfer, 110, 506-514, 2017.
13. Daniel, Y.S., Aziz, Z.A., Ismail, Z. and Salah, F., Double stratification effects
on unsteady electrical MHD mixed convection flow of nanofluid with viscous
dissipation and Joule heating, Journal of Applied Research and Technology,
15(5), 464-476, 2017.
14. Hayat, T., Qayyum, S., Khan, M.I. and Alsaedi, A., Entropy generation in
magnetohydrodynamic radiative flow due to rotating disk in presence of
viscous dissipation and Joule heating, Physics of Fluids, 30(1), 017101,
2018.
15. Hayat, T., Qayyum, S., Alsaedi, A. and Shafiq, A., Inclined magnetic field and
heat source/sink aspects in flow of nanofluid with nonlinear thermal radia-
tion, International Journal of Heat and Mass Transfer, 103, 99-107, 2016.
16. Kishan, N. and Maripala, S., Thermophoresis and viscous dissipation effects
on Darcy-Forchheimer MHD mixed convection in a fluid saturated porous
media, Advances in Applied Science Research, 3(1), 60-74, 2012.
17. Ibrahim, W., Shankar, B. and Nandeppanavar, M.M., MHD stagnation point
flow and heat transfer due to nanofluid towards a stretching sheet, Int. J. Heat
Mass Transfer., 56, 1-9, 2013.
18. Turkyilmazoglu, M., Exact analytical solutions for heat and mass transfer of
MHD slip flow in nanofluids, Chem. Eng. Sci., 84, 182-187, 2012.
19. Farooq, M., Ijaz Khan, M., Waqas, M., Hayat, T., Alsaedi, A. and Imran Khan,
M., MHD stagnation point flow of viscoelastic nanofluid with non-linear
radiation effects, Journal of Molecular Liquids., 221, 1097-1103, 2016.
20. Sheikholeslami, M., Hari Kataria, R. and Akhil Mittal, S., Effect of thermal
diffusion and heat-generation on MHD nanofluid flow past an oscillating
vertical plate through porous medium, Journal of Molecular Liquids, 257,
12-25, 2018.
21. Baag, S., Mishra, S. R., Dash, G. C. and Acharya, M. R., Entropy generation
analysis for viscoelastic MHD flow over a stretching sheet embedded in a
porous medium, Ain Shams Engineering Journal, 8(4), 623-632, 2017.
22. Sheikholeslami, M. and Rokni, H. B., Influence of melting surface on MHD
nanofluid flow by means of two-phase model, Chinese J. Physics, 55(4),
1352-1360, 2017.
23. Abdul Hakeem, A.K., Govindaraju, M. and Ganga, B., Influence of inclined
Lorentz forces on entropy generation analysis for viscoelastic fluid over a
stretching sheet with nonlinear thermal radiation and heat source/sink,
J. Heat Mass Transfer Research., 6(1), 1-10, 2019.
24. Emad M. Abo-Eladahab, Mohamed A. El Aziz, Blowing/suction effect on
hydromagnetic heat transfer by mixed convection from an inclined contin-
uously stretching surface with internal heat generation/absorption, Int. J.
Therm. Sci., 43, 709-719, 2004.
25. Subhas Abel, M. and Mahesha, N., Heat transfer in MHD viscoelastic fluid
flow over a stretching sheet with variable thermal conductivity, non-uniform
heat source and radiation, Appl. Math. Model., 32, 1965-1983, 2008.
26. Ganga, B., Govindaraju, M. and Abdul Hakeem, A.K., Effects of inclined
magnetic field on entropy generation in nanofluid over a stretching sheet
with partial slip and nonlinear thermal radiation, Iranian Journal of Science
and Technology: Transactions of Mechanical Engineering, 1139, 2018.
27. Zeeshan A, Majeed A. Effect of magnetic dipole on radiative non-Darcian
mixed convective flow over a stretching sheet in porous medium. J Nanofluids
4, 617–626, 2016.
28. H Mondal, S Ghosh, PK Roy, S Chatterjee, Effects of Ion-Slip and Hall
Currents on Magnetohydrodynamic Nanofluid Flow with Thermal Diffusion
Using Spectral Quasi-Linearization Method, Journal of Nanofluids 10 (4),
608-615.
29. S. S. Motsa. A new spectral local linearization method for nonlinear bound-
ary layer flow problems. Journal of Applied Mathematics, 2013.
16
Quickest Multi-Commodity Contraflow
with Non-Symmetric Traversal Times
Shiva Prakash Gupta1*, Urmila Pyakurel2 and Tanka Nath Dhamala2
Tri-Chandra Multiple Campus, Tribhuvan University, Kathmandu, Nepal

1
2
Central Department of Mathematics, Tribhuvan University, Kathmandu, Nepal
Abstract
Routing multiple distinct commodities from particular supply points to appro-
priate destination points via the arcs of a network topology while preserv-
ing capacity limits is the most challenging issue in operations research. The
quickest multi-commodity flow issue reduces the delivery time to complete
the process. Computationally, this problem is NP-hard. We consider asymmet-
ric transit times on anti-parallel lanes due to uneven road conditions and flow
dependency. Outbound lane capacity can be boosted by reorienting lanes towards
demand locations. We incorporate the contraflow strategy to the quickest multi-
commodity flow problem and provide an approximation scheme utilizing length-
bounded flow with non-symmetric traversal times on oppositely oriented lanes.
Keywords: Network flow, quickest multi-commodity, contraflow,

non-symmetric traversal times, length-bound
16.1 Introduction
The topology of a transport system is described as a network that cor-
relates to the transshipment of diverse commodities, with distribution
centers, demand centers, and intersections of road segments serving as
nodes and arcs serving as linkages between nodes. The starting and end-
ing destinations of commodities are known as supply points and demand
points, respectively. Flow refers to the collection of goods that have been
*Corresponding author: shivaprasadgupta99@gmail.com
239
transported via the network. In networks with temporal dimensions capac-

ity and travel times are allocated to the arcs. For more details, we refer to
[1−5].
The multi-commodity network flow problem entails transporting
several commodities from specified delivery sites to appropriate desti-
nation points while staying within arc capacity restrictions. Static flow
and dynamic flow are two types of flow that can occur in this scenario.
Maximum, maximum concurrent, and least cost flow issues are the three
categories of static flow problems. Maximum flow issues emerge when the
aggregate of all commodity flows is maximized. The maximum flow prob-
lem wherein the proportion of total demand is maximized is a maximum
concurrent flow problem.
The minimal cost flow problem is described as determining the flow
value that meets all commodity demands at the lowest cost while adhering
to capacity constraints on all arcs. Furthermore, dynamic multi-commodity
flow problems may be classified as maximal dynamic, quickest, minimum
cost, and earliest arrival flow problems (see [4−8]).
Ford and Fulkerson [9] are credited with inventing network flow over
time in the late 1950s. In the theory of supply and demand, the aim is to
compute the quickest way to fulfill the needs for something, i.e., to get it
from origin to destination is the quickest flow problem.
The first polynomial-time bound was computed by Burkard et al. [10]
for this problem using a binary search. They reduced the complexity and
gave strong polynomial-time bounds with the help of a parametric search.
The quickest transshipment problem is one in which a flow of commodities
fulfills all supplies and demands in the least amount of time.
Authors in [11] generalize the naive method to solve a dynamic flow
problem to the situation of multi-commodities. Furthermore, the authors
used length-bounded static flows instead of methodology, which used a
static minimal cost flow computation. Even though static flows do not
contain a time component, Fleischer and Skutella [11] looked at flows
that offer feasible routes using travel time as a measure of length. They
divide static flows into pathways, with each path starting at an origin node
and ending at a corresponding destination node. Multi-commodity flow
problems are more difficult than single-commodity flow problems, even
in graphs with series-parallel composition or with only two commodities.
The author in [12] found that the flow of several commodities over time
is NP-hard. NP-hardness exists in multiple products with and without
intermediate node storage for the quickest flow problem with a simple flow
path. Because this issue is NP-hard, two techniques to obtain an approxi-
mate solution have been presented in [11, 13].
Multi-commodity Contraflow with Non-symmetric Traversal Times 241
Lozovanu et al. [14] and Kappmeir [15] used a time expansion approach
for solving the maximum dynamic multi-commodity flow problem.
This approach is extended to multiple sources and a single sink multi-
commodity earliest arrival transshipment in [15].
Contraflow refers to the switching arc orientations to boost traffic flow
value and minimize traversal time by increasing the capacity of lanes in
a network. For a single-source and a single-sink, the authors in [16] pro-
posed models and strongly polynomial-time methods to solve maximum
and quickest flow problems. At time zero, these contraflows are made and
then rectified. Nath et al. [17] investigated the problem with asymmetric
traversal times on oppositely directed lanes and modified the algorithm of
[16] so that traversal time of a non-reversed arc is taken after lane reversals
and solved the problem in polynomial-time for a single-commodities. In
[18], the authors reported a polynomial-time solution to the earliest arrival
transshipment problem and the maximum flow problem with priority
on terminals. Gupta et al. [19] investigated the non-polynomial solution
of the generalized maximum flow issue on a lossy network with non-
symmetric traversal times on anti-parallel lanes. Recently, the authors in
[20] extended this to the case of multi-commodity with symmetric tran-
sit times and the same authors in [21] provided FPTAS to a maximum
dynamic multi-commodity contraflow problem with non-symmetric tra-
versal times on oppositely oriented lanes.
The fundamental objective of partial contraflow is to use the capacity
of underutilized lanes and was introduced by Pyakurel et al. [22]. In times
of emergency, the stored capacity of unoccupied arcs might be utilized for
logistical assistance and facility placement.
We present the quickest multi-commodity contraflow (QMCCF) prob-
lem with non-symmetric traversal times on anti-parallel lanes in this paper.
Fleischer and Skutella [11, 13] developed an approximation algorithm to
this problem without contraflow using a T-length bounded function.
The authors in [23, 24] introduced a partial lane reversal approach
with symmetric traversal times on oppositely directed lanes. We provide
the polynomial-time approximate solution of the problem by utilizing a
T-length bound approximation. The travel time for transshipping com-
modities from origin nodes to destination nodes is reduced by using the
lane reversal strategy in the routing problem. The study’s main relevance is
the decrease in delivery time.
The rest of the article is arranged as follows. Section 16.2 describes fun-
damental concepts with flow models. The QMCCF problem with non-
symmetric traversal times on oppositely oriented lanes is introduced in
Section 16.3. We provide a method for solving the issue in polynomial-time

in this section. The final portion concludes the paper.
16.2 Preliminaries with Flow Models

The multi-commodity flow problem entails distributing several commodi-
ties from their origin-destination pairs across a specified network to meet the
entire demand for each commodity. We define the basic notations and pro-
vide mathematical formulations for this problem, in which arc reversals are
allowed to optimize the objective value by reverting their directions as needed.
16.2.1 Mathematical Model with Contraflow

Suppose a network architecture N = (V, A, K, u, τ, di, S+, S−, T) is composed
of a collection of nodes V and arcs A. The collection of commodities is
denoted by K = {1, 2, ..., k}, with |V | = n and |A| = m. The demand di is set
and it is delivered via a particular origin-destination path si-ti, where si ∈
S+ and ti ∈ S−, ∀i ∈ K. The capacity function u : A → R+ regulates the flow
and traversal time function. τ : A → R+ computes the time it takes to carry
the flow on each arc, e. In discrete and continuous-time contexts, the time
period T is indicated as T = {0, 1, ..., T} and T = [0, T], respectively. The sets
δ+(v) = {e : e = (v, .) | . ∈ V} and δ−(v) = {e : e = (., v) | . ∈ V } represent a set
of arcs moving out from v and entering into v, respectively, so that δ+(S−) =
δ−(S+) = ∅ besides contraflow network.
The auxiliary network for network N is represented by Na = (V, Aa, K,
u , τ , di, S+, S−, T), with the edges having no direction. The arc er represents
a a
the backward arc of e. The capacity is re-defined as the total of the capaci-
ties of the anti-parallel lanes, in Na with ue = 0, if e ∉A. The traversal time
τa along with arc a in Na is the same as the traversal time of non-reversed
arcs. For a single lane, we assume τ a = τ e = τ er .. Other network parameters
are unaltered.
A static network is one that lacks a time dimension and is defined by
N = (V, A, K, u, di, S+, S−).
The static flow function is defined by x : A → R+. To compute the solu-
tion of real world dynamic flow problems, many useful properties of static
network topology are essential tools. A multi-commodity flow over time
Φ in a network N with contraflow is a collection of commodities described
by Φi : Aa × T → R+ meeting the requirements (16.2−16.4).
min T (16.1)
subject to,
T T  di if v = si
∑∑ +
θ =0 e∈δ (v )
i
Φ (θ ) −
e ∑∑
−
θ =τ e e∈δ (v )
 i
Φ (θ − τ e ) =  − di if=
ev = ti ∀i ∈ K
 0 otherwise

(16.2)
σ σ
∑∑
θ =0 e∈δ + (v )
Φie (θ ) − ∑ ∑ Φ (θ −τ ) ≤ 0, ∀v ∉ {s ,t },i ∈ K ,σ ∈T
θ =τ e e∈δ − (v )
i
e e i i
(16.3)
∑
0 ≤ Φ e (σ ) =Φie (σ ) ≤ ue + uer , ∀e ∈ Aa , i ∈ K ,σ ∈ T (16.4)
e∈δ + (v )
In this case, the third equation in constraint (16.2) denotes flow con-
servation restrictions at time T, but the constraints in (16.3) denote non-
conservation needs at intermediate time points T. Likewise, capacity with
lane reversals limits the bundle restrictions in (16.4). The purpose is to
deliver a predetermined quantity of flow to fulfill the demand for each
commodity from supply points to destination locations in the least amount
of time, as specified by the first two conditions of (16.2).
The cost is defined by
T
c(Φ) = ∑ ∑ ∑ Φ (θ )
e∈A i∈K
cei
θ=
0
i
e (16.5)
The cost bound for each commodity i is determined as:

T
∑ ∑ Φ (θ ) ≤ C
e∈A
cei
θ=
0
i
e i (16.6)
Example 1. Suppose a two-commodity network, as illustrated in Figure

16.1(a), has capacity and travel times on arcs. The pathways s1 − t1 and s2 − t2
s1 3,1 v t2 s1 3,1 v t2
2,1 4,1 4,1

3,1 5,1 4,1
4,2 3,1 7,1
x y x y
3,2 8,2
2,2 6,2(1) 4,2 10,2(1)
3,2 3,2
2,2 5,2
s2 6,1 w t1 s2 6,1 w t1
(a) (b)
Figure 16.1 (a) Network with multi-commodity having capacity and traversal time on
arcs. (b) Arc reversals with symmetric and non-symmetric traversal times.
are used to transport the first and second commodities, respectively. A contra-
flow network is depicted in Figure 16.1(b).
Authors in [25] modeled the quickest contraflow problem for a single
commodity as an integer programming problem. Furthermore, they pro-
posed greedy and bottle-neck relaxation heuristics to provide a numeri-
cal solution. Rebennack et al. [16] presented its polynomial-time solution.
However, the problem with numerous sources and/or sinks is NP-complete
via a reduction from 3-SAT [16]. In [26, 27], the continuous-time variant
of a single-source single-sink maximum and the quickest contraflow prob-
lems are polynomially solvable. The authors in [22] introduced and solved
the problem with partial lane reversals.
16.3 QMCCF with Non-Symmetric Transit Times

The lane reversal approach for the QMCCF problem with non-symmetric
traversal times on oppositely oriented lanes is discussed in this section. In
a network architecture where arc reversals are allowed, a solution to this
problem meets given demands at defined nodes in the shortest time feasible.
We present a polynomial-time approximation scheme (PTAS) to com-
pute the solution. Furthermore, our technique may be used to save lanes
that do not need to be reversed to shorten the time. Instead of symmetric
traversal times on oppositely oriented lanes in [23, 24], we consider asym-
metric traversal times on oppositely oriented lanes and present Problem 1.
Problem 1. Assume a network N = (V, A, K, u, τ, S+, S−, di, T) has non-

symmetric traversal times on oppositely oriented lanes. The QMCCF problem
aims to find the minimum feasible time that meets the requirements (16.2),
(16.3), and (16.4) to transport a specified number of demands di, ∀i ∈ K from
supply points to corresponding destination points at a bounded cost by revert-

ing the orientation of only essential arcs from the beginning.
Even with a series-parallel network composition or network having two
commodities only, the multi-commodity flow over time problem is NP-hard
[12]. The proof is based on reductions from the NP-hard PARTITION
and 3-PARTITION issues. In addition, the problem of maximal multi-
commodity flow with no limits on intermediate node storage is NP-hard. A
time expansion is a tool that provides the pseudo-polynomial time solution.
The same hardness exists to the quickest multi-commodity flow prob-
lem having a simple path and without intermediary node storage accord-
ing to [12].
Authors in [25] used 3-SAT to show that lane reversal is NP-complete.
As a result, we have Theorem 1.
Theorem 1. The QMCCF problem with bounded cost having asymmetric

traversal times on anti-parallel arcs is NP-hard.
As the quickest multi-commodity flow problem is NP-hard, two tech-
niques are proposed to provide an approximate solution in [13]. One is the
discretization of bigger time steps as opposed to single time steps, while
the other is the length-bounded flow.
The authors in [23, 24] incorporated partial lane reversals with equal traversal
times on oppositely oriented lanes and provided the solution using the same
approach. We consider non-symmetric traversal times on oppositely oriented
lanes and provide a PTAS to Problem 1 with the help of length-bounded flow.
16.3.1 Approximation Approach for the QMCCF

A single or multi-commodity path flow is considered as a length-bounded
flow, where each used path is subjected to a length restriction. Let Pi rep-
resent the sum of all si − ti routes ∀i ∈ K in the network Na. A static flow
x is known as a T-length bounded if every constituent flow xi for each
i ∈ K can be broken down into the amount of flows x Pi , i.e., x i = ∑
x Pi
∑
P ∈Pi
with x Pi > 0 and the length τ P = τ e of any path P ∈ Pi is at most T.
e ∈P
A T-length, restricted static flow x that matches multi-commodity con-
straints is NP-hard to discover [13]. This issue, on the other hand, is
polynomial-time approximated.
We compute an approximate solution to the QMCCF problem with
non-symmetric traversal times on anti-parallel lanes by applying length-
bounded flows as described [23, 24] and presented in Algorithm 1. In Step
2 of our approach, we first review the solution technique of [11, 13]. All
other steps of the algorithm are feasible according to [23]. The removal of
the cycle flow in Step 3 suggests that flow only travels in one direction, not
both. In Step 5, the capacity of the unutilized arcs is preserved. We argue
that a possible approximate solution to the QMCCF with non-symmetric
traversal time on oppositely oriented lanes on network N would likewise
be a feasible approximate solution to the quickest multi-commodity flow
issue on modified network Na.
The authors in [17] have shown that the complexity of the lane rever-
sal problems having non-symmetric travel times on oppositely oriented
lanes is the same as symmetric travel time on oppositely oriented lanes.
As a result, an approximate QMCCF with non-symmetric traversal time
on oppositely oriented lanes may be optimally calculated on network N.
Hence, we have Theorem 2.
Algorithm 1 T-length Bound Algorithm for QMCCF with Asymmetric

Transit Times
Input: A network N = (V, A, K, u, τ, di, S+, S−, T ) with non-symmetric

travel times
Output: The QMCCF
1. Construct the modified network by adding capacities of

anti-parallel arcs on Na = (V, Aa, K, ua, τa, di, S+, S−, T) as
u=
a ue + uer ,,
τ e if arc e re is
if arc r reverted towards e
is reverted towards e
τ a : = τa: = 
τ er if arc e eisreverted
if arc reverteetowards er er .
d towards
2. Calculate the flow on the modified network Na by applying

an algorithm of [13] to the problem having bounded cost.
3. Remove cycles by decomposing the flow along the ori-
gin-destination pathways and cycles ∀i ∈ K.
4. Revert er ∈ A to the capacity xe − ue iff xe > ue, ue substituted
k k
by 0 whenever e ∉A, ∀i, where x e = ∑

i =1
x ei and ue = ∑u
i =1
i
e
5. For each e ∈ A, if er is reverted, the saved capacity of arc er is

ua − xe and the saved capacity of arc e is zero. If none of the
arcs are e and er is reversed, sc(e) = ue − xe > 0.
Theorem 2. An approximate solution to the QMCCF problem with non-

symmetric traversal times on oppositely oriented lanes can be determined in
polynomial-time by using T-length bound.
Example 2. Consider a network as illustrated in Figure (16.1)(a) with two

commodities having demands d1 = 12 and d2 = 8. The quickest time with-
out contraflow with symmetric and asymmetric traversal times on lanes can
be computed by using Figure 16.1(a), and the quickest time with contraflow
having symmetric and asymmetric traversal times on lanes can be computed
by using Figure 16.1(b), respectively.
There is only one path for commodity-1 from s1 to t1 and only one path for
commodity-2, i.e., s2 to t2 before contraflow with symmetric or asymmetric
traversal times (cf. Figure 16.1(a)). A 5-length bound is required for com-
puting the solution of Problem 1 to the quickest multi-commodity without
contraflow having symmetric traversal times (cf. Figure 16.1(a)), and it is
repeated four times. As a result, T1 = 8 units of time are required to meet the
indicated demand d1 = 12 of commodity-1. Similarly, a 5-length bound is
required and is repeated four times for commodity-2. As a result, it takes T2 =
8 units of time to meet demand d2 = 8 of commodity-2. Hence, the shortest
possible time needed to fulfill both needs is T = max{T1, T2} = {8, 8} = 8. If we
take asymmetric traversal times on oppositely oriented lanes before contra-
flow (cf. Figure 16.1(a)), a 4-length bound is essential to meet the demand of
commodity-1 and it is repeated 4-times. So, T1 = 7 is the shortest time needed
to fulfill the demand of commodity-1. Similarly, a 4-length bound is essential
to meet the need of commodity-2 and it is repeated 4-times. So, to fulfill the
need of commodity-2, T2 = 7 is the quickest time. Hence, the shortest possible
time needed to fulfill both needs is T = max{T1, T2} = {7, 7} = 7.
However, if contraflow is used (see Figure 16.1(b)), and traversal time is
symmetric, the time taken to meet both needs is T = 6 units.
In the same way, if lane reversal is used (see Figure 16.1(b)), and traversal
time is non-symmetric on oppositely oriented lanes, it requires T = 5 units
of time to meet both demands. Thus, contraflow with symmetric traversal
times saves approximately 25 percent of the time, whereas contraflow with
non-symmetric traversal times on oppositely oriented lanes saves approxi-
mately 28.7 percent of the time.
Table 16.1 and Table 16.2 summarize the data obtained from Example 2.
Table 16.1 Quickest time without contraflow.

LB without contraflow ST LB without contraflow AT
8 7
Table 16.2 quickest time with contraflow.

LB with contraflow ST LB with contraflow AT
6 5
LB=length bound, ST=symmetric transit times, AT=asymmetric
transit times.
16.4 Conclusions
One of the primary issues in operations research is routing many com-
modities from their delivery points to target points over a shared network.
A crucial concern is the reduction of time (cost). A well-known quickest
flow problem was addressed to deliver the requisite demand in the shortest
feasible period. For a single-commodity, this issue was solved in strongly
polynomial-time, but for several commodities it is NP-hard. However, a
polynomial-time approximation technique based on a length-bounded
function was found. The lane reversals technique is an essential tool for
improving the quickest time in the two-way network. We incorporated this
approach in length-bounded approximation.
In this work, we investigated the QMCCF problem with non-symmetric
traversal times on oppositely oriented lanes. We presented its mathemati-
cal model and provided a polynomial-time approximation algorithm. This
study minimizes the routing time significantly. By analyzing the results
of the previously studied problem with constant transit time, it would be
interesting to extend these strategies to flow-dependent transit times. The
findings in this research have both theoretical and practical implications.
Acknowledgments
The first author wishes to thank the University Grant Commission of
Nepal for the Ph.D. research grant and the second author wishes to thank
the Alexander Von Humboldt Foundation for the Digital Cooperation
Fellowship (August 1, 2021 to January 31, 2022).
References
1. R. K. Ahuja, T. L. Mangati, and J. B. Orlin, Network flows: theory, algorithm,
and applications, Prentice Hall, Englewood Cliffs, 1993.
2. A. Assad, Multi-commodity network flows a survey, Networks, Vol. 8(1), pp.

37-91, 1978.
3. J. Kennington, A survey of linear cost multi-commodity network flows,
Operations Research, Vol. 26(2), pp. 209-236, 1978.
4. K. Salimifard and S. Bigharaz, The multi-commodity network flow problem:
state of the art classification, applications, and solution methods, Springer,
pp. 1-47, 2020.
5. I.-L. Wang, Multi-commodity network flows: A survey, part I: Applications
and formulations, International Journal of Operations Research, Vol. 15(4),
pp. 145-153, 2018.
6. A. Ali, R. Helgason, J. Kennington, and H. Lall, Computational compari-
son among three multi-commodity network flow algorithms, Operations
Research, Vol. 28(4), pp. 995-1000, 1980.
7. J. A. Tomlin, Minimum cost multi-commodity network flows, Operational
Research, Vol. 14, pp. 45-51, 1966.
8. U. Pyakurel, S. P. Gupta, D. P. Khanal, and T. N. Dhamala, Efficient algorithms
on multi-commodity flow over time problems with partial lane reversals,
International Journal of Mathematics and Mathematical Sciences, Hindawi,
pp. 1-13, 2020. Article ID 2676378, https://doi.org/10.1155/2020/2676378.
9. L. R. Ford and D. R. Fulkerson, Flows in networks, Princeton University
Press, Princeton, New jersey, 1962.
10. R. E. Burkard, K. Dlaska, and B. Klinz, The quickest flow problem. ZOR-
Methods and Models of Operations Research, Vol. 37, pp. 31-58, 1993.
11. L. Fleischer and M. Skutella, The quickest multicommodity flow problem,
In International Conference on Integer Programming and Combinatorial
Optimization, Springer, pp. 36-53, 2002.
12. A. Hall, S. Hippler, and M. Skutella, Multi-commodity flows over time: effi-
cient algorithms and complexity. Science Direct, Vol. 379, pp. 387-404, 2007.
13. L. Fleischer and M. Skutella, Quickest flows over time, SIAM Journal on
Computing, Vol. 36(6), pp. 1600-1630, 2007.
14. D. Lozovanu and M. Fonoberova, Optimal dynamic multicommodity flows
in networks, Electron Notes Discrete Math, Vol. 25, pp. 93-100, 2006.
15. P. W. Kappmeier, Generalizations of flows over time with application in evacu-
ation optimization, PhD Thesis, Technical University, Berlin, Germany, 2015.
16. S. Rebennack, A. Arulselvan, L. Elefteriadou, and P. M. Pardalos, Complexity
analysis for maximum flow problems with arc reversals, Journal of
Combinatorial Optimization, Vol. 19, pp. 200-216, 2010.
17. H. N. Nath, U. Pyakurel, and T. N. Dhamala, Network reconfiguration with
orientation dependent transit times, International Journal of Mathematics
and Mathematical Sciences, Hindawi, pp. 1-11, 2021. Article ID 6613622,
https://doi.org/10.1155/2021/6613622.
18. S. P. Gupta, U. Pyakurel, and T. N. Dhamala, Network flows with arc rever-
sals and non-symmetric transit times, American Journal of Mathematics and
Statistics, Vol. 11(2), pp. 27-33, 2021.
19. S. P. Gupta, U. Pyakurel, and T. N. Dhamala, Generalized dynamic contra-

flow with non-symmetric transit times, American Journal of Computational
and Applied Mathematics, Vol. 11(1), pp. 12-17, 2021.
20. S. P. Gupta, U. Pyakurel, and T. N. Dhamala, Multi-commodity flow problem
on lossy network with partial lane reversals. Annals of Operations Research
(ANOR), 323(1-2), 45-63, 2023, https://doi.org/10.1007/s10479-023-05210-y
21. S. P. Gupta, U. Pyakurel, and T. N. Dhamala, Dynamic multi-commodity
contraflow problem with asymmetric transit times, Journal of Applied
Mathematics, Hindawi, pp. 1-8, 2022. Article ID 3697141, https://doi.
org/10.1155/2022/3697141.
22. U. Pyakurel, H. N. Nath, S. Dempe, and T. N. Dhamala, Efficient dynamic
flow algorithms for evacuation planning problems with partial lane reversal,
Mathematics, Vol. 7, pp. 1-29, 2019.
23. T. N. Dhamala, S. P. Gupta, D. P. Khanal, and U. Pyakurel, Quickest multi-
commodity flow over time with partial lane reversals, Journal of Mathematics
and Statistics, Vol. 16(1), pp. 198-211, 2020.
24. S. P. Gupta, D. P. Khanal, U. Pyakurel, and T. N. Dhamala, Approximate algo-
rithms for continuous-time quickest multi-commodity contraflow problem,
The Nepali Mathematical Sciences Report, Vol. 37(1-2), pp. 30-46, 2020.
25. S. Kim, S. Shekhar, and M. Min, Contraflow transportation network recon-
figuration for evacuation route planning, IEEE Trans. Knowl. Data Eng., Vol.
20, pp. 1115-1129, 2008.
26. U. Pyakurel and T. N. Dhamala, Evacuation planning by earliest arrival con-
traflow, Journal of Industrial and Management Optimization, Vol. 13, pp. 487-
501, 2017.
27. U. Pyakurel, T. N. Dhamala, and S. Dempe, Efficient continuous contraflow
algorithms for evacuation planning problems. Annals of Operations Research
(ANOR), Vol. 254, pp. 335-364, 2017.
17
A Mathematical Representation
for Deteriorating Goods with a
Trapezoidal-Type Demand, Shortages
and Time Dependent Holding Cost
Ruma Roy Chowdhury *
Department of Basic Science and Humanities, University of Engineering &

Management, Kolkata, India
Abstract
A production inventory model for a perishable good decaying at a constant rate
has been devised in this research article. A ramp-type demand function has been
considered, that is, the demand follows an accelerated growth initially for some
time and finally becomes constant. The model allows shortage and is completely
backlogged. The cost of holding the inventory is considered to vary linearly with
time. Four cases have been developed depending on the position of time point
at which the demand becomes constant in a trapezoidal-type demand. The pro-
duction cycle restarts after a certain time. Optimal production stopping time and
resuming time are calculated to optimize the expense of the production set-up per
unit. The model is illustrated with a numerical.
Keywords: Trapezoidal demand, holding cost varying with time, completely

backlogged short-ages, constant deterioration
17.1 Introduction
Demand is always fluctuating according to the requirements and prefer-
ences of customers and is quite unpredictable. In case of new launches
of items like cosmetics, electronic gadgets with new technologies, trendy
Email: ruma roych@yahoo.co.in
251
garments, etc. in the market, the items are in high demand initially. But,
with the passage of time as more improvised items come into the market,
the demand for the old ones become constant and stabilized. In order to
deal with this kind of demand trend, trapezoidal demand has been dealt
with in various studies since Ritchie (1980) [1]. Shaikh et al. (2020) [2]
devised a stock optimization model for perishing goods with a preser-
vation facility and a trapezoidal demand. Yang (2019) [3] developed an
optimization model for a trapezoidal order trend having bi-level business
credit type financing while placing orders. Chauhan et al. (2021) [4] devel-
oped an EOQ model for a trapezoidal type ordering system. During the
stock out or shortage duration, backlogging is considered, also allowing
certain discount strategies. Sana (2010) [5] considered deterioration and
partial backlogging. Kawakatsu (2010) [6] devised an optimal refilling
strategy for a retailer. A seasonal good with a trapezoidal order trend is
considered with focus on Special Display Goods. Avinadav et al. (2013) [7]
have developed a model where the demand varies with price of the item.
Wang and Huang (2014) [8] devised a model for periodic items having
price and trapezoidal-type order trends and deterioration. Ahmed et al.
(2013) [9] proposed a fresh method for investigating the EOQ (Economic
Order Quantity) assuming a trapezoidal demand, partial backlogging, and
a general perish rate.
Perishing of items is an unavoidable phenomenon that should not be
neglected while formulating an inventory system. But, it definitely depends
on the kind of item taken into consideration. For example, for highly
deteriorating items like fruits, vegetables, etc., deteriorating items should
be taken as a variable function of time, quantity, etc. But, items such as
metals (steel utensils, iron rods, etc.), medicines (expiry), etc. deteriorate
extremely slow and hence the rate of deterioration might be considered
constant. In this study, the items being studied are of the non-depreciat-
ing type like electronic goods, medicines, COVID protective gears, etc.
In recent time, Halima et al. (2021) [10] devised an overtime production
model for goods with a constant rate of perish and the demand depend-
ing on non-linear price and stock dependent order trend. Mandal and Pal
(1998) [11] have developed inventory models for perishing goods that fol-
lows a trapezoidal demand.
Shortages of items are very common in the market system. When short-
ages occur, it is completely a customer’s will whether to wait for the back-
order or to go to some other seller for instant availability. Backlogging
rate varies with the holdback time for the next refilling. Goyal (1988) [12]
established a model for replenishment of trendy/fashionable inventories
including shortages. Giri et al. (2000) [13] devised a model on lot sizing
Mathematical Representation for Deteriorating Goods 253
study for perishing goods with stockout situations and demand depending
on time. Wang (2002) [14] devised a replenishment strategy for perishing
items with inventory running with shortages and thus, partial backlog-
ging. Ghosh and Chaudhuri (2005) [15] developed an economic order
quantity model for a perishable good with trended demand and stockout
situations in each cycle. Pentico and Drake (2011) [16] did a study on
deterministic models for the economic order quantity models and eco-
nomic production quantity models considering backordering. Sharma
and Singh (2013) [17] devised a replica for perishable goods, accounting
for consumption rate and backlogging. Roy Chowdhury et al. (2014) [18]
designed an order-level inventory model for a decaying good having a
quadratic type ordering pattern and backlogging during a stock out situ-
ation in every cycle. Tripathi (2015) [19] developed a model for perishing
goods with an escalating order trend and stock out situation under infla-
tionary spiral.
In the present work, a production-inventory model for a constantly per-
ishable good with a trapezoidal-type order trend has been considered. The
production starts at an initial time and is proportional to demand. The cost
of holding inventory varies linearly with time. Shortage is allowed and is
completely backlogged. Three cases have been dealt with depending on the
position of the time at which the trapezoidal-type demand gains stability.
We have illustrated the model numerically.
17.2 Assumptions and Notations

17.2.1 Assumptions
1. A single item is considered

2. Order for the item is supposed to be a ramp-type
3. The holding expense varies linearly with time, that is Ch = h +
st(h, s > 0)
4. The production rate P is proportional to demand
5. Shortages occur and are completely backlogged
6. The stock is subjected to constant deterioration, θ
7. The time horizon, T, for the model is finite
8. The perished items are neither repaired nor replaced
9. Lead time is negligible
17.2.2 Notation
A: The unit ordering cost per cycle ($/unit)
P (t): Production rate varies directly with demand D(t), that is, P (t) =
βD(t); this means that the higher the demand, the greater the production;
this is required for ramp-type demand as initially demand is more, hence
production should also be more to satisfy the customer demand, but the
demand stabilizes after sometime and the production has to be reduced
accordingly in order to avoid wastage of items and loss
ν: The parameter of the trapezoidal demand function (time point) where
demand becomes constant
D(t): In case of certain items like COVID gear, with the outbreak of
COVID-19 in India in the year 2020, the demand of masks and PPEs saw
an increasing demand in the beginning but after some time the demand
became stabilized. This kind of demand is a ramp type demand which can
be mathematically expressed as: Demand at the time t ≥ 0
D(t) = a + bt − (t − ν)H(t − ν) (17.1)
where H(t − ν) is a Heviside function defined as
H (t − v ) = 1 (t ≥ ν )
(17.2)
= 0, (t < ν )
CO : Overall ordering expense

CH : Overall holding expense
CP : Overall production expense
CD : Overall deterioration expense
CS : Overall shortage expense
ch : Time dependent Holding Cost (h+st) in the time interval [0, t2]
cp : Unit production expenses per item per unit time ($/unit/unittime)
cd : Expenses of deterioration per item per unit time ($/unit)
cs : Shortage expenses per unit per unit time
θ : Constant deterioration rate of the item, where 0 < θ << 1
T : Total time period of the inventory cycle
ATC : Optimum aggregate total expense per unit time of the production-
inventory system calculated using the optimized values of manufacturing
end time, the instant when the stock exhausts, and the production resum-
ing time
ta : Time instant when the production ceases

tb : Time instant when the inventory level gets exhausted in an unpro-
ductive state
tc : Time at which the production starts again
17.2.3 Decision Variables

tam : Optimal time when the production ceases
tbm : Optimal time when the inventory level exhausts in an
unproductive state
tcm : Optimal time at which the production starts again
17.3 Formulation and Solution

The present article develops a production inventory system for a perish-
able item decaying at a constant rate. The cycle begins with no items in
the inventory, that is, the stock level is zero. The production P initiates
at time t = 0 and pauses at time t = ta. Demand is ramp-type and starts
along with production. Within the time interval, 0 ≤ t ≤ ta, the inventory
builds up at a rate governed by the rate of production, the rate at which
the orders are placed, and constant deterioration θ. The deterioration
is acting on the accumulating inventory. In the interval ta ≤ t ≤ tb, the
demand and the constant deterioration cause the inventory to deplete
down to zero level at t = tb. After this point, the shortage is allowed and
is completely backlogged. As a result, the demand accumulates until t =
tc. Here, the production restarts to satisfy the backlogged demand as well
as meet the current demand. ta, tb, and tc are the decision variables. The
cycle recurs after period T.
Depending on the position of the time point µ on the time horizon
(0 < t < T ), we consider the following three cases:
i) 0 < ν < ta < tb < tc < T

ii) 0 < ta < ν < tb < tc < T
iii) 0 < ta < tb < ν < tc < T
17.3.1 Case 1 i) 0 < ν < t1 < t2 < t3 < T

The formulation of the production model is as follows:
dI1A (t )
= (β − 1)(a + bt ) − θ I1A (t ), (0 ≤ t ≤ ν )
dt
dI1B (t )
= (β − 1)(a + bν ) − θ I1B (t ), (ν ≤ t ≤ t a )
dt
dI1C (t )
= −(a + bν ) − θ I1C (t ), (t a ≤ t ≤ tb )
dt
dI1D (tt )
= −(a + bν ), (tb ≤ t ≤ t c )
dt
dI1E (t )
= (β − 1)(a + bν ), (t c ≤ t ≤ T ) (17.3)
dt
with boundary conditions I1A(0) = 0, I1A(ν) = I1B(ν), I1B(ta) = I1C(ta), I1C(tb) =

0, I1D(tb) = 0, and I1E(T ) = 0.
The above differential equations have been solved to get the following
expressions:
(β − 1)(aθ + b)(tθ − 1)) −tθ (β − 1)(aθ − b)

I1A (t ) = −e , (0 ≤ t ≤ ν )
θ2 θ2
(β − 1)(a + bν ) −tθ (a + bν ) tbθ
I1B (t ) = +e (e − β e tα θ ), (ν ≤ t ≤ t a )
θ θ
(a + bν ) (tb −t )θ
I1C (t ) = (e − 1), (t a ≤ t ≤ tb )
θ
I1D (t=) (a + bν )(tb − t ), (tb ≤ t ≤ t c )
t ) (bν − a)(β − 1)(t − T ),
I1EE (= (t c ≤ t ≤ T )
(17.4)
The total expenditure of the production inventory system comprises of:
1. Ordering Cost (CO) = A.

2. Holding Cost
CH= (h + st )Z0v I1A (t )dt + (h + st )Zvta I1B (t )dt + (h + st )Zttab I1C (t )dt
(17.5)
3. Shortage Cost
C=
S {
cs −Zttbc I1D (t )dt − ZtTc I1E (t )dt } (17.6)
4. Production Cost
C=
P c p {−Z0v β (a + bt )dt + Zvta β (a + bν )dt + ZtTc β (a + bν )dt} (17.7)
5. Deterioration Cost = (Production - Demand) in the time

period (0, tb)
CD cd {(β − 1)Z0v (a + bt )dt + (β − 1)Zvta (a + bν )dt − Zttab (a + bν )dt }

=
(17.8)
17.3.2 Case 2 ii) 0 < ta < ν < tb < tc < T

The formulation of the inventory model is as follows:
dI 2 A (t )
= (β − 1)(a + bt ) − θ I 2 A (t ), (0 ≤ t ≤ t a )
dt
dI 2 B (t )
= −(a + bt ) − θ I 2 B (t ), (t a ≤ t ≤ ν )
dt
dI 2C (t )
= −(a + bν ) − θ I 2C (t ), (ν ≤ t ≤ tb )
dt
dI 2 D (t )
= −(a + bν ),
= (tb ≤ t ≤ t c )
dt
dI 2 E (t )
= (β − 1)(a + bν ), (t c ≤ t ≤ T ) (17.9)
dt
with boundary conditions being I2A(0) = 0, I2A(ta) = I2B(ta), I2B(ν) = I2C(ν),

I2C(tb) = 0, I2D(tb) = 0, I2E(T ) = 0,
expressions:
(β − 1)(aθ + b(−1 + tθ ))
I 2 A (t ) =
θ2 (0 ≤ t ≤ t a )
(β − 1)(aθ − b)
−e −tθ ,
θ2
(aθ + b(−1 + tθ ))
I 2 B (t ) = −
θ2
{
+ e −tθ
(aθ + b(−1 + t aθ )) taθ (β − 1)(aθ − b)
θ2
βe −
θ2
, } (t a ≤ t ≤ ν )
(a + bν ) (tb−t )θ
I 2C (t ) = (e − 1), (ν ≤ t ≤ tb )
θ
I 2 D (t ) = (a + bν )(tb − t ), (tb ≤ t ≤ t c )
I 2 E (t=) (a + bβ )(β − 1)(t − T ), (t c ≤ t ≤ T )
(17.10)

2. Holding Cost
CH= (h + st )Z0ta I 2 A (t )dt + (h + st )Ztav I 2 B (t )dt + (h + st )Zvtb I 2C (t )dt

(17.11)
3. Shortage Cost
C=
S {
cs −Zttbc I 2 D (t )dt − ZtTc I 2 E (t )dt } (17.12)
4. Production Cost
C=
P c p {−Z0td β (a + bt )dt + ZtcT β (a + bν )dt} (17.13)

period (0, tb)
CD cd {(β − 1)Z0td (a + bt )dt − Ztdv (a + bν )dt − Zvtb (a + bν )dt }

=
(17.14)
17.3.3 Case 3 iii) 0 < ta < tb < ν < tc < T

The formulation of the above production inventory model is as follows:
dI 3 A (t )
= (β − 1)(a + bt ) − θ I 3 A (t ), (0 ≤ t ≤ t a )
dt
dI 3 B (t )
= −(a + bt ) − θ I 3 B (t ), (t a ≤ t ≤ tb )
dt
dI 3C (t )
= −(a + bt ), (tb ≤ t ≤ ν )
dt
dI 3D (t )
= −(a + bν ), (ν ≤ t ≤ t c )
dt
dI 3 E (t )
= (β − 1)(a + bν ), (t c ≤ t ≤ T ) (17.15)
dt
Boundary conditions: I3A(0) = 0, I3A(ta) = I3B(ta), I3B(tb) = 0, I3C(tb) = 0,

I3C(ν) = I3D(ν), I3E(T) = 0,
expressions:
(β − 1)(aθ + b(−1 + tθ )) −tθ (β − 1)(aθ − b)

I 3 A (t ) = −e , (0 ≤ t ≤ t a )
θ2 θ2
(aθ + b(−1 + tθ )) (tb −t )θ (aθ + b(−1 + θ tb )
I 3 B (t ) = − +e , (t a ≤ t ≤ tb )
θ2 θ2
b
I 3C=
(t ) a(tb − t ) + (tb2 − t 2 ), (tb ≤ t ≤ v )
2
b
I 3D (t ) = −(a + bv )t + atb + (tb2 + v 2 ), (v ≤ t ≤ t c )
2
I 3 E (t=
) (a + bv )(β − 1)(t − T ), (t c ≤ t ≤ T )
(17.16)

2. Holding Cost
CH = (h + st )Z td I 3 A (t )dt + (h + st )Z tb I 3 B (t )dt (17.17)

0 ta
3. Shortage Cost
∫ {a(t − t ) + 2 (t − t )}dt
v
b 2 2
C=
s cs {− b b
tb
− {−(a + bv )t + at + (t + v )}dt
tc
b
∫ v 2
b b
2 2
(17.18)
T
∫
− {(a + bv )(β − 1)(t − T )}dt }
tc
4. Production Cost
{
CP = c p − J t0a β (a + bt )dt + J T β (a + bν )dt
tc } (17.19)

period (0, tb)
{
CD = cd ( β − 1) J t0a (a + bt )dt − J tb (a + bt )dt
td } (17.20)
The average total system expenditure for the above three instances is
given by:
TC = T1 [CO + CH + CS + CD + CP ]
The total system cost TC depends on three variables ta, tb, and tc. We
need to find the optimized values of ta, tb, and tc that is, tm, tm, and tm respec-
tively, that minimize the average total system cost.
∂TC ∂TC ∂TC
tm, tm, and tm are the solution of = 0, = 0, = 0, respectively,
∂t1 ∂t 2 ∂t 3
which are the necessary conditions for the optima. The sufficient condi-
tions that minimize the average total cost are the Hessian Matrix H > 0 at
tm, tm, tm, where
 ∂ 2 ATC ∂ 2 ATC ∂ 2 ATC 

 *2 
 ∂t1 ∂t1*∂t 2* ∂t1*∂t 3* 
 ∂ 2 ATC ∂ 2 ATC ∂ 2 ATC 
H= * * (17.21)
 ∂t 2∂t1 ∂t 2*2 ∂t 2*∂t 3* 
 ∂ 2 ATC ∂ 2 ATC ∂ 2 ATC 
 ∂t *∂t * ∂t 3*∂t 2* ∂t 3*2 
 3 1
17.4 Numerical Example

Example 1: Case 1 (0 < µ < ta < tb < tc < T)
We now illustrate the above developed production inventory model with
a numerical example. Random data taken are as follows: A = 100, Cd = 10,
a = 70, b = 20, β = 0.001, θ = 0.001, Cs = 15, T = 8, ν = 5.6, Cp = 10, e =
2.718, h = 5, s = 0.05 (with appropriate units).
We use MATHEMATICA 7.0 to calculate the optimized values
m m m
tam , tbm , tcm . The optimized values of ta, tb, tc are t a = 7.2, t b 7.38, t c 7.6 units
and the optimized average expense is ATC = 15, 613 units.
Example 2: Case 2 (0 < ta < µ < tb < tc < T)

a = 70, b = 20, β = 0.001, θ = 0.001, Cs = 15, T = 8, ν = 5.7, Cp = 10, e =
2.718, h = 4, s = 0.04 (with appropriate units).
m m m
tam , tbm , tcm . The optimal values of ta, tb, tc are ta = 5.618, tb 8.05, tc = 8.08
units and the optimized average expense is ATC = 16, 041 units.
Example 3: Case 3 (0 < ta < tb < µ < tc < T )

a = 80, b = 10, β = 0.001, θ = 0.001, Cs = 20, Ch = 4, T = 8, ν = 7.42, Cp =
10, e = 2.718, h = 5, s = 0.04 (with appropriate units).
m m m
tam , tbm , tcm . The optimal values of ta, tb, tc are ta = 7.2954, tb 7.3939, tc = 7.496
units and the optimized average expense is ATC = 1, 47, 896 units.
17.5 Discussion
Case 1: (0 < µ < ta < tb < tc < T)
Here, ν < ta and this implies that the demand curve becomes constant
before the production stops (ta = 7.2). Hence, the manufactured items will
take some time to sell and the inventory level will fall to zero after some
delay. This aspect is clear from the optimum values received in Example 1.
Case 2: (0 < ta < µ < tb < tc < T)

Here, ta < ν < tb and this implies that the demand curve becomes constant
after the production stops (ta = 5.618). After the stabilization of demand,
the level of the inventory reduces at a very slow rate. Hence, the level
reduces to zero after a long duration at tb = 8.05. This is clear from the opti-
mum values of tb, tc received in Example 2.
Case 3: (0 < ta < tb < µ < tc < T)

Here, tb < ν and this implies that the demand curve is stabilized after the
inventory level dips below the zero mark and the shortage begins. Because
the inventory is running in shortage, therefore the optimal time to start the
production is almost immediately. This aspect is clear from the optimum
values in Example 3.
17.6 Inference
This study deals with a production-reserve (stock) model with a trapezoi-
dal-type demand for an item that perishes constantly and inventory that is
stored at a flexible cost. Shortages have been considered. The items consid-
ered are the ones like COVID gear (masks, PPE kits, etc.). Three cases have
been discussed related to the demand change time and the inventory levels
and production start and stop times. The decision variables are the pro-
duction halt/break off time (ta), the time at which the stock gets exhausted
(tb), and production restart time (tc). In the future, this study can be taken
further by considering a flexible deterioration rate.
References
1. E. Ritchie Practical inventory replenishment policies for a linear trend in
demand followed by a period of steady demand J. Oper. Res. Soc., 31 (7)
(1980), pp. 605-613.
2. An inventory model for deteriorating items with preservation facility of ramp
type demand and trade credit Ali Akbar Shaikh, Gobinda Chandra Panda,
Md. Al-Amin Khan, Abu Hashan Md. Mashud, Amiya Biswas https://doi.
org/10.1504/IJMOR.2020.110895 International Journal of Mathematics in
Operational Research Volume 17, Issue 4, 2020.
3. Open Journal of Business and Management, Vol.7 No.2, April 2019 An
Inventory Model for Ramp-Type Demand with Two-Level Trade Credit
Financing Linked to Order Quantity Hui-Ling Yang DOI: 10.4236/
ojbm.2019.72029
4. Chauhan et al. [2021] An order quantity scheme for ramp type demand
and back-logging during stock out with discount strategy, Anand Chauhan,
Shilpy Tayal, https://doi.org/10.1504/IJSOI.2021.114113 International

Journal of Services Operations and Informatics Volume 11, Issue 1
5. Sana, S., (2010), ‘Optimal selling price and lotsize with time varying deterio-
ration and partial backlogging’, Appl. Math. Comput., Vol. 217, pp. 185-194.
6. H. Kawakatsu, Optimal retailers replenishment policy for seasonal products
with ramp-type demand rate, IAENG Int. J. Appl. Math. 40 (4) (2010) 17.
7. Avinadav, T., Herbon, A., Spiegel, U. 2013. Optimal inventory policy for a
perishable item with demand function sensitive to price and time. Int. J.
Production Economics 144, 497506.
8. Pricing for seasonal deteriorating products with price-and ramp-type
time-dependent demand C Wang, R Huang - Computers & Industrial
Engineering, 2014 - Elsevier
9. M.A. Ahmed, T.A. Al-Khamis, L. Benkherouf Inventory models with ramp
type demand rate, partial backlogging and general deterioration rate Appl.
Math. Comput., 219 (2013), pp. 4288-4307.
10. Alexandria Engineering Journal Volume 60, Issue 3, June 2021, Pages 2779-
2786 An overtime production inventory model for deteriorating items with
nonlinear price and stock dependent demand Mohammad Abdul Halima
A.Paul MonaMahmoud B.Alshahranic Atheelah Y.M.Alazzawi Gamal
M.Ismaile
11. Mandal, B., Pal, A.K., 1998. Order level inventory system with ramp type
demand rate for deteriorating items. Journal of Interdisciplinary Mathematics
1 (1), 4966.
12. Goyal, S.K., (1988), ‘A heuristic for replenishment of trended inventories con-
sidering shortages’, Journal of the Operational Research, Vol.39, pp.885-887.
13. Giri, B.C., Chakraborty, T., Chaudhuri, K.S., (2000), ‘A note on a lot sizing
heuristic for deteriorating items with time-varying demands and shortages’,
Computers and Operations Research, Vol.27, pp.495-505.
14. Wang, S.P., (2002), ‘An inventory replenishment policy for deteriorating
items with shortages and partial backlogging’, Computers & Operations
Research, Vol.29(14), pp.2043-2051.
15. Ghosh, S.K., Chaudhuri, K.S., (2005), ‘An EOQ model for a deteriorating
item with trended demand and variable backlogging with shortages in all
cycles’, Int. J. Adv. Model. Optim. (Buchar. Rom.), Vol.7(1), pp.57-68.
16. Pentico, D.W., Drake, M.J., (2011), ‘A survey of deterministic models for the
EOQ and EPQ with partial backordering’, European Journal of Operations
Research, Vol. 214(2), pp. 179-198.
17. Sharma, S., Singh, S.R., (2013), ‘An inventory model for decaying items, con-
sidering multi variate consumption rate with partial backlogging’, Indian
Journal of Science and Technology, Vol. 6(7), pp.4870-4880.
18. Roy Chowdhury, R., Ghosh, S.K., Chaudhuri, K. S., (2014), ‘An Order-Level
Inventory Model for a Deteriorating Item with Time-Quadratic Demand
and Time-Dependent Partial Back-logging with Shortages in All Cycles’,
American Journal of Mathematical and Management Sciences, vol.33(2),
pp.75-97.
19. Tripathi, R.P., (2015), ‘Economic order quantity for deteriorating item with
nondecreasing demand and shortages under inflation and time discounting’,
International Journal of Engineering, Vol.28(9), pp.1295-1302.
18
An Amended Moth Flame Optimization
Algorithm Based on Fibonacci
Search Approach for Solving
Engineering Design Problems
Saroj Kumar Sahoo* and Apu Kumar Saha
Department of Mathematics, National Institute of Technology Agartala,

Tripura, India
Abstract
The moth flame optimization (MFO) algorithm is a swarm intelligence (SI) based
algorithm which gained popularity among researchers due to a special kind of
movement mechanism, namely, a transverse orientation mechanism of the moth
in nature. Like other SI based algorithms, it also suffers from good quality solution
and slow convergence speed. To avoid the drawbacks, a new variant of MFO algo-
rithm, namely, a Fibonacci technique based MFO algorithm (in short Ft-MFO)
is presented in this paper. We merged the concept of Fibonacci search method
in the classical MFO algorithm to improve the search quality and accelerate the
convergence speed of the MFO algorithm. To validate the performance of the pro-
posed algorithm, Ft-MFO is compared with six popular stochastic optimization
algorithms on an IEEE CEC2019 test suite and two constraint engineering design
problems. Experimental results demonstrate that the proposed Ft-MFO algorithm
is superior to the other stochastic algorithms in terms of solution quality and con-
vergence rate.
Keywords: Moth flame optimization algorithm, Fibonacci search method,
benchmark functions
*Corresponding author: sarojlipu.gugle@gmail.com
265
18.1 Introduction
The challenge of finding the better solution in optimization problems is an
interesting topic of research due to its importance in both academia and
industry. If the number of optimization parameters increases, the optimi-
zation problem complexity also increases. In recent decades, researchers are
very much interested in machine learning and artificial intelligence (AI) tech-
niques as real-life difficulties, such as constrained or unconstrained, linear or
nonlinear, continuous or discontinuous, that can be easily tackle by AI and
machine learning techniques [1, 2]. Due to the aforementioned character-
istics, there are various levels of difficulty to handle such types of difficulties
using conventional techniques through numerical or mathematical program-
ming, namely quasi-Newton approach, quadratic programming, conjugate
gradient, fast steepest method [3], etc. In various existing research [4], it has
been experimentally proved that the above-mentioned approaches are not
efficient enough to handle non-differentiable and real-life multimodal issues.
In contrast, the nature-based algorithms have played a substantial role in
tackling these issues, as these algorithms are simple and can be easily applied.
A few of them are the Genetic Algorithm (GA) [5], Particle Swarm Optimizer
(PSO) [6], Differential Evolution (DE) [7], Spotted Hyena Optimization
Algorithm (SHO) [8], Butterfly Optimization Algorithm (BOA) [9], Whale
Optimization Algorithm (WOA) [10], Sine Cosine Algorithm (SCA) [11],
Moth Flame Optimizer (MFO) [12], Salp Swarm Algorithm (SSA) [13], etc.
Typically, these algorithms initiate by a set of randomly chosen initial solu-
tions and continue until they get the optimal solution of a problem. When the
algorithm exceeds the pre-determined number of iterations it will automati-
cally be terminated. There is an increasing interest for the efficient, low-cost,
and effective implementation of such metaheuristic algorithms these days.
The MFO algorithm is the subject of this article. In 2015, Mirjalili dis-
covered MFO, a swarm intelligence-based algorithm. Transverse orienta-
tion is used by moths to navigate in the wild and served as an inspiration
for MFO. Spiral flight search and simple flame generation (SFG) are two
of the most important MFO tactics. In the SFG technique, the best moths
and flames collected thus far can be used to directly manufacture flames.
By following moths’ transverse direction, the SFS technique allows moths
to spiral toward the flames to update the positions in an iterative way. It is
possible that MFO will find the best option in the available search space.
The transverse orientation of moths in particular is critical to the effective-
ness of MFO. The key advantage of MFO on other algorithms is its abil-
ity to tackle several tough issues involving confined and unknown search
spaces such as optical network unit placement [14], automatic generation
An Amended Moth Flame Optimization Algorithm 267
control problem [15], image segmentation [16], feature selection [17],

medical diagnoses [18], and smart grid system [19].
As a new population-based optimization method, MFO’s performance still
needs to be improved and studied in some dimensions, including convergence
speed and global search capabilities. Various academics have come up with a
variety of ways to fix the MFO algorithm’s flaws and a few of these are pre-
sented here. In order to tackle the shortcomings of the MFO method, Hongwei
et al. [20] presented a new form of the algorithm called chaos-enhanced MFO,
which incorporates a chaotic map. Yueting et al. [21] proposed a series of
new MFO algorithm variants by integrating MFO with Gaussian, Cauchy,
and Levy mutations to reduce the disadvantages of the MFO and expand the
search capability of MFO. To achieve a more stable balance between diversity
and intensification in the MFO algorithm by embedding Gaussian mutation
and chaotic local search, Xu et al. [22] developed CLSGMFO. Three additional
adjustments were introduced to the MFO and suggested E-MFO by Kaur et al.
[23] to keep a favorable balance between diversification and intensification
and boost exploration and exploitation, respectively. For the prediction of
software errors, Tumar et al. [24] implemented a modified MFO method and
presented an extended binary moth-flame optimization algorithm (EBMFO).
To strike a compromise between global and local search capabilities, Wei Gu
and Gan Xiang [25] suggested a new modified MFO algorithm termed as a
“multi operator MFO algorithm” (MOMFO). An updated version of the MFO
algorithm was created by Ma and colleagues [26] to alleviate the MFO’s short-
comings, including delayed and local minimum convergence. A new version
of the MFO, namely EMFO, based on the mutualism phase of symbiotic
organism search, has been proposed by Sahoo et al. [27]. Few recent upgraded
versions of MFO algorithms are discussed in [28–34].
Researchers have also developed more efficient algorithms in addition to the
ones listed above. For example, Chakraborty et al. [35] introduced WOAmM,
where the authors embedded the modified mutualism phase in WOA to
alleviate inherent drawbacks of WOA. Nama et al. [36] proposed the hybrid
SOS (HSOS) by integrating Simple Quadratic Interpolation (SQI) with SOS
to enhance the robustness of the process. An effective hybrid method called
m-MBOA has been developed by Sharma and Saha [37]. Because of this, BOA’s
overall performance was enhanced by using mutualism in the exploring part.
Chakraborty et al. [38] introduced an efficient hybrid method called HSWOA
by hybridizing the HGS algorithm into the WOA algorithm and applied it
to solve different engineering design problems. Sharma et al. [39] introduced
a different type of modification in BOA named mLBOA in which Lagrange
interpolation and SQI are used in exploration and exploitation phases, respec-
tively, to improve the original BOA algorithm. In the present study, a modified
MFO is formulated, namely Ft-MFO with the help of the Fibonacci technique.
The major steps involved in this work are as follows:
• Firstly, a non-linear function is embedded into the classical

MFO algorithm to maintain exploration and exploitation
capability of the suggested Ft-MFO.
• Secondly, we merged the concept of Fibonacci search method
in MFO to boost the solution quality of the proposed Ft-MFO.
• The efficiency of the new Ft-MFO is examined with six pop-
ular stochastic optimization algorithms on an IEEE CEC
2019 test suite and two constrained engineering problems.
• A Friedman rank test was carried out in order to investigate
the performance of the newly suggested Ft-MFO algorithm.
The rest of this article is structured as follows. Section 18.2 provides an

overview of the MFO algorithm. Section 18.3 shows the suggested Ft-MFO
algorithm. Section 18.4 presents the simulation outcomes and perfor-
mance metrics. Statistical tests and a convergence analysis are discussed
in in Section 18.5, as well as real life problems. Finally, conclusions with
future enhancements are discussed in Section 18.6.
18.2 Classical MFO Algorithm

This section presents the origin of the MFO algorithm and its working
process with the mathematical formulation presented below.
Moths are members of the Arthropoda family. Moth navigation tech-
niques are one-of-a-kind, which draws researchers’ attention. In the sub-
sections that follow, the MFO algorithm is represented mathematically.
The positions of all moths are represented using a vector of choice vari-
ables. Take a peek at the moth matrices below.
 x1,1 x1,2  x1,n −1 x1,n 

 X1   
   x2,1    x 2,n 
X2
X= =       (18.1)
    
   x N −1,1    x N −1,n 
 X N  
 x N ,1 x N ,2  x N ,n −1 x N ,n 
where Xi = [ xi ,1 , xi ,2 ,…, xi ,n ] , i ∈{1,2,…, N }.

N is the number of moths in the original population, while n denotes

variable numbers. The flame matrix (FMx) is the second crucial factor of
the MFO algorithm. Mathematically, it can be represented as follows:
 Fm1,1 Fm1,2  Fm1,n −1 Fm1,n 

 FMx1   
   Fm2,1    Fm2,n 
FMx = 
FMx = 
2
    

    
   FmN −1,1    FmN −1,n 
 FMx N  
 FmN ,1 FmN ,2  FmN −1 FmN ,n 
(18.2)
Moths move spirally when they are nearer to the flame, therefore the
author used a logarithmic spiral function which is as follows:
 δ ⋅ e bt ⋅ cos(2π t ) + Fm (k ), i ≤ N .FM
K +1 i i
x i = bt (18.3)
 δ i ⋅ e ⋅ cos(2π t ) + FmN . FM (k ), i ≥ N .FM
where = δ i xiK − Fmi represents the distance of the moth at ith place and
its specific flame (Fmi) and t can be any random number between −1 and 1.
Here, b is a fixed constant used to recognize the spiral flight shape. A moth
moves like a helix towards the flame with a one-dimensional approach and
a discrete value of t and represented as follows:
 −1 
a1 = −1 + current iter   (18.4)
 maxiter 
t = (a1 – 1) × r + 1 (18.5)
where maxiter and a1 indicate the maximum iterations and the constant of
convergence, respectively, which decreases linearly from (−1) to (−2). In
every iteration, flame positions for the current and last iterations are col-
lected and arranged as per the fitness value for the global and local search.
The number of flames (N.FM) that have been lowered over the iteration
can be calculated using the formula below.
 (N .FM Lst it − 1) 
=N .FM round  N .FM Lst it − crnt .it  (18.6)
 max it 
18.3 Proposed Method

The main characteristics of generalized optimization algorithms include
exploration and exploitation. Exploration involves searching the entire
search region. However, exploitation is characterized as examining limited
areas within a large search field. These effects contribute to the algorithm’s
ability to prevent local minima stagnation (as a result of exploitation) and
to promote convergence and solution variety (from exploration). The equi-
librium between these two occurrences is also vital. Any algorithm that
achieves these three criteria is deemed as a state-of-the-art algorithm.
In MFO, the spiral motion of moths around the flame provides explora-
tion and exploitation. It is easier to understand exploration and exploitation
when the exponent factor ‘t’ is used to explain it. Iteratively, in traditional
Ft-MFO, the parameter t is taken from the linearly decreasing range of (−1)
to 2, but in our new approach we have introduced a non-linear decreasing
range of (−1) to 2, which helps maintain an equilibrium of global and local
search by first exploring the search space and then gradually shrinking and
exploiting the region found. The following is a mathematical formula for
the parameter ‘r.’
3
 −iteration 
 
r = −2 + e  k ×maxiter 
(18.7)
where k is a constant and its value is 0.55, which was suitably chosen so that
it helps in both global and local searches and is represented in Figure 18.1.
Fibonacci Search Method (FSM)

The FSM is a mathematical process that shifts and narrows down the search
range by using Fibonacci numbers to obtain the extreme value of uni-
modal functions. The optimal point is always contained within the range
being narrowed. Shifting can take place in both directions. The values of
that function at two experiment points determine the changing direction.
The Fibonacci numbers, which are defined as follows, are the foundation
of the FSM.
Fib = [F1, F2, F3, … .., Fn], where, Fi ∀ i = 1,2,..n are Fibonacci numbers
and generated by the following equation:
F0 = 1 = F1, Fno = Fno−1 + Fno−2, no = 2,3,4, … . n (18.8)

Parameter (r) vs Iterations

-1
-1.1
MFO Ft-MFO
-1.2
-1.3
-1.4
Parameter (r)
-1.5
-1.6
-1.7
-1.8
-1.9
-2
0 100 200 300 400 500 600 700 800 900 1000
Number of iterations
Figure 18.1 Non-linear adaption curve.
Let t1 and t2 be two experimental numbers of any finite length of inter-

val with upper and lower limits as UL and LL, respectively. Calculate two
initial points:
F  F 
t1 LL +  no − 2  (UL − LL), t 2 =
= UL −  no − 2  (UL − LL) (18.9)
 Fno   Fno 
The range is moved to the right because the function’s value at t2 is

greater than that at t1 and to the left if t1 is greater than that at t2. The new
value of t3 and t4 are generated using the Fibonacci search formula as t3 = t1
and t4 = t2. If two functional values are unequal, then only one (t3 or t4) will
be considered as a new experimental point, whereas the other will be the
same as either of t1 and t2 depending on the contracting direction. In the
case of two equal function values, then both t3 and t3 form new experimen-
tal points. Due to the high computational efficiency of FSM, the author
of [40] used a modified Fibonacci search method for the partially shaded
solar PV array. Recently, Yazici et al. [41] applied a modified Fibonacci
search method for conversion systems of wind energy. We embedded the
concept of the Fibonacci search method after the position update phase
of MFO. The pseudocode of the projected Ft-MFO method is presented
below as Algorithm 18.1.
.
Algorithm 18.1: Pseudocode of Ft-MFO Algorithm
Objective function f(X), X = (X1 X2……....Xdimension);

for i = 1: No. of search agents
for j = 1: dimension
Generate n organism solutions Xi(i = 1, 2, … ., n) using equation
X (i, j) = L .B(i) + (U .B (i) – LB(i)) *rand
end
end
While Current iteration < Maximum iteration
if Iteration == 1
Enter N .FM = N in initial population
else
Employ Eq. (18.6).
end
FM = Fitness Function f(X);
if Iteration == 1
Sort the moths according to FM
Update the Flames
Iteration = 0;
else
Sort the moths based on FM from last iteration
Update the Flames
end
Reduce the converge constant
for j = 1: No. of search agents
for k = 1: dimension
Find r and t using Eq. (18.5) & (18.7)
Update moths position as to their particular flame using Eq. (18.3)
end
end
Apply Fibonacci search process using Eq. (18.8) and Eq. (18.9)
Current iteration = Current iteration + 1
End while
Output: The best solution with the minimum fitness
18.4 Results and Discussions on IEEE CEC 2019

Benchmark Problems
It should be noted that the code for the proposed Ft-MFO algorithm has
also been written in the MATLAB environment and run on a 1.80 GHz i5
8th generation computer with 8.00GB of RAM and MATLAB R2015a. In
this paper, the proposed Ft-MFO has been applied on IEEE CEC 2019 test
problems and the performance results are compared with a wide variety of
state-of-the-art methods. The selected parameters of all population-based
optimization algorithms involve the maximum iteration and number of
populations as 1000 and 30, respectively. To determine the best set of
parameters for Ft-MFO and all other algorithms, 30 trials are performed
for each possible set of parameters. The studied cases are presented below.
The CEC 2019 benchmark function has stored more complex type func-
tions than other CEC test suits. The author [42] developed some complex
single objective optimization problems, namely, “The 100-digit Challenge”.
There is a total of ten number functions that are multimodal and insepara-
ble in nature. The first three functions (F1-F3) of CEC 2019 have different
search ranges and different dimensions but the other seven functions (F4-
F10) have the same search range with the same dimensions. The optimal
value of all ten complex functions (F1-F10) is set to one.
Table 18.1 shows the performance results of 10 (ten) IEEE CEEC’2019
test suites using Ft-MFO and six basic current state-of-the-art optimization
techniques, including MFO, DE, SOS, JAYA, BOA, and WOA. Table 18.2
shows how many times the efficiency of Ft-MFO is higher, similar to, and
worse than that of other methods. In 8, 7, 8, 9, 7, and 8 benchmark functions,
respectively, Ft-MFO outperforms MFO, DE, SOS, JAYA, BOA, and WOA;
comparable results are seen in 2, 0, 0, 0, 2, and 0 benchmark functions,
while poorer results are shown in 0, 3, 2, 1, 1, and 2 benchmark functions.
The Friedman Test, introduced by Milton Friedman [43], is a non-para-
metric statistical test. It is used to spot treatment variations across sev-
eral test runs. The Friedman rank test has been employed in this study to
equate the mean results of the algorithms for each benchmark issue. Table
18.3 shows that Ft-MFO has the lowest rank which is highlighted in bold,
implying that its performance is superior to other algorithms. The conver-
gence performance of Ft-MFO with other algorithms is shown in Figure
18.2, which is very competitive with respect to other methods.
Table 18.1 Simulation results of Ft-MFO and other algorithms on CEC’2019 suite.
Algorithm F1 F2 F3 F4 F5
AVG STD AVG STD AVG STD AVG STD AVG STD
Ft-MFO 1 0 5 0 8.40 1.59 3.34e+01 3.39e+01 1.64 1.68
MFO 1 0 5 0 9.62 9.60e-01 1.05e+02 1.56e+01 1.00e+02 2.59e+01
DE 1.82E+06 8.25E+05 1.93e+03 3.40e+02 4.58 6.11e-01 8.54 1.99 1.77 2.64e-02
SOS 4.75E+03 6.88E+03 2.65e+02 1.96e+02 4.61 9.78e-01 2.08e+72 3.42e+72 1.19e+73 4.40e+73
JAYA 6.53E+06 2.83e+06 4.68e+03 7.78e+02 8.09 7.57e-01 4.10e+01 4.90 2.60 1.67e-01
BOA 1 0 5 0 6.79 1.15 1.26e+02 2.15e+01 1.09e+02 2.33e+01
WOA 1.45E+08 1.17E+08 1.10E+04 3.36E+03 8.16 1.60 6.62E+01 2.40E+01 3.83 1.43
AVG STD AVG STD AVG STD AVG STD AVG STD
Ft-MFO 3.07 1.21 8.22e+02 2.23e+02 4.25 2.41e-01 1.26 9.43e-02 2.15e+01 1.75e-01
(Continued)
Table 18.1 Simulation results of Ft-MFO and other algorithms on CEC’2019 suite. (Continued)
MFO 1.21e+01 1.09 2.29e+03 2.51e+02 5.13 1.79e-01 3.86 6.23e-01 2.16e+01 1.43e-01
DE 1.11 1.48e-01 1.07e+03 1.24e+02 4.48 2.71e-01 1.29 4.20e-02 2.25e+01 2.52
SOS 2.15 2.22e-01 4.55e+71 4.09e+71 5.50 0 1.59e+71 2.34e+71 2.18e+01 8.74e-04
JAYA 5.90 9.23e-01 1.51e+03 2.09e+02 4.28 2.08e-01 1.56 1.11e-01 2.16e+01 1.27e-01
BOA 1.36e+01 1.15 2.07e+03 2.89e+02 5.12 2.83e-01 4.14 5.51e-01 2.16e+01 1.85e-01
WOA 9.94 1.84 1.49E+03 2.85E+02 4.73 3.09E-01 1.48 1.69E-01 21.46 1.12E-01
Table 18.2 Experimental results of Ft-MFO with other algorithms on IEEE

CEC’2019 test suite.
MFO DE SOS JAYA BOA WOA
Superior to 8 7 8 9 7 8
Similar to 2 0 0 0 2 0
Inferior to 0 3 2 1 1 2
Table 18.3 Friedman rank test.

Algorithm Mean rank Rank
Ft-MFO 3.40 1
SOS 3.76 2
BOA 4.08 3
MFO 4.66 4
DE 4.69 5
WOA 4.67 6
JAYA 4.79 7
Converges graph of function number 5 Converges graph of function number 6

6 3
DE
DE Ft-MFO
5 Ft-MFO 2.5 JAYA
JAYA BOA
BOA 2 WOA
log(f(x)-f(x*))
4 WOA
log(f(x)-f(x*))
MFO
MFO SOS
SOS 1.5
3
1
2
0.5
1
0
0 0 200 400 600 800 1000 1200 1400 1600 1800 2000
0 200 400 600 800 1000 1200 1400 1600 1800 2000 Function Evaluation
Function Evaluation
Converges graph of function number 7 Converges graph of function number 8

8 1.8
DE DE
Ft-MFO 1.7 Ft-MFO
JAYA
7.5 JAYA
1.6 BOA
BOA
1.5
log(f(x)-f(x*))
WOA WOA
MFO
log(f(x)-f(x*))
MFO
7 SOS 1.4 SOS
1.3
6.5 1.2
1.1
6 1
0.9
0 200 400 600 800 1000 1200 1400 1600 1800 2000
5.5 Function Evaluation
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Function Evaluation
Figure 18.2 Convergence graph of Ft-MFO with other competative algorithms.

18.5 Real-Life Applications

Optimal gas production capacity and three-bar truss issues are utilized as
real-world problems to show how well the proposed Ft-MFO algorithm
works on real-life problems.
18.5.1 Optimal Gas Production Capacity Problem

The above problem is an unconstrained problem and details with mathe-
matical representation are elaborated in [44].
The simulation results of the above problem are presented in Table 18.4.
In this table, the results of DE, GSA, and DE-GSA are taken from [45].
From Table 18.4, we can conclude that the performance of the proposed
Ft-MFO method is shown to be superior to that of other methods.
18.5.2 Three-Bar Truss Design (TSD) Problem

The challenge of designing a three-bar truss is one of structural opti-
mization in civil engineering. Complex limited search space makes this
challenge useful [46]. Two design criteria, buckling stress and deflection,
have been modified in order to produce the lowest possible weight. The
multiple components of the three-bar truss design issue are depicted in
Figure 18.3.
Our new Ft-MFO algorithm is used to solve the TSD problem, which
is then compared to the CS, DEDS, PSO-DE, Tsa, and MBA algorithms
taken from the literature [12]. The comparative findings, as well as optimal
weights and optimal variables, are provided in Table 18.5. Our suggested
Ft-MFO technique outperforms the other three algorithms, as shown in
Table 18.5.
Table 18.4 Simulation results for optimal gas production problem.

Item DE MFO GSA DE-GSA Ft-MFO
x1 17.5 17.5 17.5 17.5 17.5
x2 600 600 600 600 600
f(x) 169.844 71.4495 169.844 169.844 71.4459
1 2 3 D
D
A
4 A
Figure 18.3 Three-bar truss design problem.
Table 18.5 Simulation results for three-bar truss problem.

Algorithm Optimal variables Optimal weight
x1 x2
Ft-MFO 0.408966 0.288146 174.2762166
MFO 0.788244770931922 0.788244770931922 263.895979682
CS 0.78867 0.40902 263.9716
DEDS 0.78867513 0.40824828 263.8958434
PSO-DE 0.7886751 0.4082482 263.8958433
Tsa 0.788 0.408 263.68
MBA 0.7885650 0.4085597 263.8958522
18.6 Conclusion with Future Studies

This paper presents an upgraded variety of the classic MFO algorithm,
namely an amended MFO (Ft-MFO) which uses a Fibonacci search con-
cept and a non-linear adaption formula to improve the MFO algorithm and
make a good tradeoff between diversification and intensification. To eval-
uate the performance of Ft-MFO, IEEE CEC 2019 benchmark functions
have been considered for experimentations and compared with the basic
MFO, DE, JAYA, BOA, SOS, and WOA. The Friedman Test is used to mea-
sure the effectiveness of the suggested Ft-MFO algorithm. It has also been
used to solve two engineering issues, providing a better result than previ-
ous algorithms to validate the proposed Ft-MFO. According to simulation
results, using the global best solution in the optimization process quickly
brought the proposed method into focus. To avoid local optima traps and
premature convergence, the proposed technique is helpful.
In the future we can extend it to multi-objective optimization, apply it
to solve higher constraint optimization problems like car-side crash prob-
lems, robot gripper problems, welded beam design problems, etc. We can
generate an efficient metaheuristic algorithm by hybridizing our suggested
approach, Ft-MFO, with any other meta-heuristic algorithm.
References
1. Abbassi, R., Abbassi, A., Heidari, A. A., & Mirjalili, S. (2019). An efficient
salp swarm-inspired algorithm for parameters identification of photovoltaic
cell models. Energy Conversion and Management, 179, 362–372.
2. Faris, H., Ala’M, A.-Z., Heidari, A. A., Aljarah, I., Mafarja, M., Hassonah, M.
A., & Fujita, H. (2019). An intelligent system for spam detection and identi-
fication of the most relevant features based on evolutionary random weight
networks. Information Fusion, 48, 67–83.
3. McCarthy, J. F. (1989). Block-conjugate-gradient method. Physical Review
D, 40(6), 2149.
4. Wu, G., Pedrycz, W., Suganthan, P. N., & Mallipeddi, R. (2015). A variable
reduction strategy for evolutionary algorithms handling equality constraints.
Applied Soft Computing, 37, 774–786.
5. Holland, J. H. (1992). Genetic algorithms. Scientific American, 267(1),
66–73.
6. Kennedy, J., Eberhart, R. Particle swarm optimization. In: Proceedings of
ICNN’95—international conference on neural networks, Perth, Australia,
1995, 1942–1948.
7. Storn, R., & Price, K. (1997). Differential evolution–a simple and efficient
heuristic for global optimization over continuous spaces. Journal of Global
Optimization, 11(4), 341–359.
8. Dhiman, G., & Kumar, V. (2017). Spotted hyena optimizer: A novel bio-
inspired based metaheuristic technique for engineering applications.
Advances in Engineering Software, 114, 48–70.
9. Arora, S., & Singh, S. (2015). Butterfly algorithm with levy flights for
global optimization. 2015 International Conference on Signal Processing,
Computing and Control (ISPCC), 220–224.
10. Mirjalili, S., & Lewis, A. (2016). The whale optimization algorithm. Advances
in Engineering Software, 95, 51–67.
11. Mirjalili, S. (2016). SCA: A sine cosine algorithm for solving optimization
problems. Knowledge-Based Systems, 96, 120–133.
12. Mirjalili, S. (2015). Moth-flame optimization algorithm: A novel nature-

inspired heuristic paradigm. Knowledge-Based Systems, 89, 228–249.
13. Mirjalili, S., Gandomi, A. H., Mirjalili, S. Z., Saremi, S., Faris, H., & Mirjalili,
S. M. (2017). Salp Swarm Algorithm: A bio-inspired optimizer for engineer-
ing design problems. Advances in Engineering Software, 114, 163–191.
14. Singh, P., & Prakash, S. (2019). Optical network unit placement in Fiber-
Wireless (FiWi) access network by Whale Optimization Algorithm. Optical
Fiber Technology, 52, 101965.
15. Mohanty, B. (2019). Performance analysis of moth flame optimization algo-
rithm for AGC system. International Journal of Modelling and Simulation,
39(2), 73–87.
16. Khairuzzaman, A. K. M., & Chaudhury, S. (2020). Modified MothfFlame
optimization algorithm-based multilevel minimum cross entropy thresh-
olding for image segmentation: International Journal of Swarm Intelligence
Research, 11(4), 123–139.
17. Gupta, D., Ahlawat, A. K., Sharma, A., & Rodrigues, J. J. P. C. (2020). Feature
selection and evaluation for software usability model using modified moth-
flame optimization. Computing, 102(6), 1503–1520.
18. Muduli, D., Dash, R., & Majhi, B. (2020). Automated breast cancer detection
in digital mammograms: A moth flame optimization-based ELM approach.
Biomedical Signal Processing and Control, 59, 101912.
19. Kadry, S., Rajinikanth, V., Raja, N. S. M., Hemanth, D. J., Hannon, N. M., &
Raj, A. N. J. (2021). Evaluation of brain tumor using brain MRI with modi-
fied-moth-flame algorithm and Kapur’s thresholding: A study. Evolutionary
Intelligence, 1–11.
20. Hongwei, L., Jianyong, L., Liang, C., Jingbo, B., Yangyang, S., & Kai, L.
(2019). Chaos-enhanced moth-flame optimization algorithm for global
optimization. Journal of Systems Engineering and Electronics, 30(6),
1144–1159.
21. Xu, Y., Chen, H., Luo, J., Zhang, Q., Jiao, S., & Zhang, X. (2019). Enhanced
moth-flame optimizer with mutation strategy for global optimization.
Information Sciences, 492, 181–203.
22. Xu, Y., Chen, H., Heidari, A. A., Luo, J., Zhang, Q., Zhao, X., & Li, C. (2019).
An efficient chaotic mutative moth-flame-inspired optimizer for global opti-
mization tasks. Expert Systems with Applications, 129, 135–155.
23. Kaur, K., Singh, U., & Salgotra, R. (2020). An enhanced moth flame optimi-
zation. Neural Computing and Applications, 32(7), 2315–2349.
24. Tumar, I., Hassouneh, Y., Turabieh, H., & Thaher, T. (2020). Enhanced binary
moth flame optimization as a feature selection algorithm to predict software
fault prediction. IEEE Access, 8, 8041–8055.
25. Gu, W., & Xiang, G. (2021). Improved moth flame optimization with mul-
tioperator for solving real-world optimization problems. 2021 IEEE 5th
Advanced Information Technology, Electronic and Automation Control
Conference (IAEAC), 5, 2459–2462.
26. Ma, L., Wang, C., Xie, N., Shi, M., Ye, Y., & Wang, L. (2021). Moth-flame
optimization algorithm based on diversity and mutation strategy. Applied
Intelligence.
27. Sahoo, S. K., Saha, A. K., Sharma, S., Mirjalili, S., & Chakraborty, S. (2022).
An enhanced moth flame optimization with mutualism scheme for function
optimization. Soft Computing, 26(6), 2855–2882.
28. Sahoo, S. K., & Saha, A. K. (2022). A hybrid moth flame optimization
algorithm for global optimization. Journal of Bionic Engineering, 19(5),
1522–1543.
29. Sahoo, S. K., Saha, A. K., Nama, S., & Masdari, M. (2022). An improved moth
flame optimization algorithm based on modified dynamic opposite learning
strategy. Artificial Intelligence Review, 1–59.
30. Sahoo, S. K., Saha, A. K., Sharma, S., Mirjalili, S., & Chakraborty, S. (2022).
An enhanced moth flame optimization with mutualism scheme for function
optimization. Soft Computing, 26(6), 2855–2882.
31. Sahoo, S.K., Saha, A.K., Ezugwu, A.E. et al. Moth flame optimization: theory,
modifications, hybridizations, and applications. Arch Computat Methods
Eng (2022). https://doi.org/10.1007/s11831-022-09801-z.
32. Chakraborty, S., Saha, A. K., Sharma, S., Sahoo, S. K., & Pal, G. (2022).
Comparative performance analysis of differential evolution variants on engi-
neering design problems. Journal of Bionic Engineering, 19(4), 1140–1160.
https://doi.org/10.1007/s42235-022-00190-4.
33 Sahoo, S. K., Sharma, S., & Saha, A. K. (2023). A Novel variant of moth flame
optimizer for higher dimensional optimization problems, 1–27.
34 Sahoo, S. K., & Saha, A. K. (2022, August). A modernized moth flame optimi-
zation algorithm for higher dimensional problems. In ICSET: International
Conference on Sustainable Engineering and Technology, (Vol. 1, No. 1, pp.
9–20).
35. Chakraborty, S., Kumar Saha, A., Sharma, S., Mirjalili, S., & Chakraborty, R.
(2021). A novel enhanced whale optimization algorithm for global optimiza-
tion. Computers & Industrial Engineering, 153, 107086.
36. Nama, S., Saha, A. K., & Ghosh, S. (2017). A hybrid symbiosis organisms
search algorithm and its application to real world problems. Memetic
Computing, 9(3), 261–280.
37. Sharma, S., & Saha, A. K. (2020). m-MBOA: A novel butterfly optimiza-
tion algorithm enhanced with mutualism scheme. Soft Computing, 24(7),
4809–4827.
38. Chakraborty, S., Saha, A. K., Chakraborty, R., Saha, M., & Nama, S. (2022).
HSWOA: An ensemble of hunger games search and whale optimization algo-
rithm for global optimization. International Journal of Intelligent Systems,
37(1), 52–104.
39. Sharma, S., Chakraborty, S., Saha, A. K., Nama, S., & Sahoo, S. K. (2022).
mLBOA: A Modified Butterfly Optimization Algorithm with Lagrange
Interpolation for Global Optimization. Journal of Bionic Engineering.
40. Ramaprabha, R. (2012). Maximum power point tracking of partially shaded

solar PV system using modified Fibonacci search method with fuzzy con-
troller. 12.
41. Yazıcı, İ., Yaylacı, E. K., Cevher, B., Yalçın, F., & Yüzkollar, C. (2021). A new
MPPT method based on a modified Fibonacci search algorithm for wind
energy conversion systems. Journal of Renewable and Sustainable Energy,
13(1), 013304.
42. Price, K. V., Awad, N. H., Ali, M. Z., & Suganthan, P. N. (2018). Problem
definitions and evaluation criteria for the 100-digit challenge special session
and competition on single objective numerical optimization. In Technical
Report. Nanyang Technological University.
43. Milton F. (1937). The use of ranks to avoid the assumption of normal-
ity implicit in the analysis of variance. Journal of the American Statistical
Association. 32: 675–701.
44. Nama, S. (2021). A modification of I-SOS: performance analysis to large
scale functions. Applied Intelligence, 51(11), 7881–7902.
45. Muangkote, N., Sunat, K., & Chiewchanwattana, S. (2016). Multilevel thresh-
olding for satellite image segmentation with moth-flame based optimization.
2016 13th International Joint Conference on Computer Science and Software
Engineering (JCSSE), 1–6.
46. Gandomi A.H., Yang X.S., Alavi A.H. (2011). Mixed variable structural opti-
mization using firefly algorithm. Computers & Structures 89:2325-36.
19
Image Segmentation of Neuronal Cell
with Ensemble Unet Architecture
Kirtan Kanani1, Aditya K. Gupta1, Ankit Kumar Nikum1*,
Prashant Gupta2 and Dharmik Raval1
1
Department of Mechanical, Sardar Vallabhbhai National Institute of Technology,
Surat, Gujarat, India
2
Licious, Banglore, India
Abstract
Medical image segmentation consists of heterogeneous pixel intensities, noisy/
ill-defined borders, and high variability, which are significant technical obstacles
for segmentation. Also, generally the requirement of annotated samples by the
networks is significantly large to achieve high accuracy. Gathering this dataset for
the particular application and annotating new images is both time-consuming and
costly. Unet solves this problem by not requiring vast datasets for picture segmen-
tation. The present work describes the use of a network that depends on augmen-
tation of the existing annotated dataset to make better use of these examples and
a comparison of encoder accuracy on Unet is presented. The encoder principal
function is to reduce image dimensionality while keeping as much information
as possible. EfficientNets tackles both of these issues and utilizing it as an encoder
of Unet can further enhance its accuracy. The test dataset highest F1-Score and
IoU were 0.7655 and 0.6201 on neuronal data values, respectively. It outperforms
Inception and ResNet encoder networks with considerably more parameters and
a higher inference time.
Keywords: Image segmentation, computer vision, deep learning, neural cell
*Corresponding author: nikumankit@gmail.com
283
19.1 Introduction
Neurodegenerative illnesses cover a diverse spectrum of ailments result-
ing from progressive damage to cells and nervous system connections.
These illnesses eventually lead to diseases such as Alzheimer’s and
Parkinson’s disease [1], which are among the six highest causes of death.
According to the WHO report [2] on neurological disorders, neuro-
logical illnesses account for 12% of all fatalities worldwide. Figure 19.1
shows the architecture of the modified U-Net model used for training.
The goal of finding treatments and solutions for neurodegenerative ill-
nesses is becoming more urgent with the increase in population. In 2015,
neurological illnesses caused 250.7 million disability-adjusted life years
(DALYs) and 9.4 million deaths, accounting for 10.2% of global DALYs
and 16.8% of global deaths [3]. Our work is focused on assisting doctors
in determining the effects of medical treatments in neuron cells using
machine learning and computer vision.
For biomedical image analysis, the segmentation of 2D pictures is a
critical problem [4]. We have proposed an Unet based neural network
that upgrades the FCN model for biomedical application, i.e., RIC
Unet, for nuclei segmentation. Following Unet architecture, RIC-Unet
uses a channel attention mechanism, residual blocks, and multi-scale
ENCODER DECODER
I O
N U
P T
U P
T (16x16x2048)
U
(8x8x1824)
(16x16x3104)
T
(16x16x768) (16x16x1056) (32x32x640)
(32x32x256) (64x64x128) (64x64x368)
(32x32x384)
(64x64x240) (128x128x64) (128x128x208)
(256x256x32)
(128x128x144)
(128x128x48)
Concatenate
Figure 19.1 Architecture of Unet with EfficentNet as encoder. The decoder is similar to the
original Unet model but with cropping and contamination from various encoder elements.
Image Segmentation of Neuronal Cell with Unet Architecture 285
techniques for segmentation of nuclei accurately. This network is supe-

rior to other techniques in terms of analysing indicators and cost-
effectiveness in assisting doctors in diagnosing the details of these
histological images.
In many medical picture segmentation issues U-Net provides a
state-of-the-art performance. U-Net with residual blocks or blocks
with dense connections, Attention U-Net, and recurrent residual con-
volutional U-Net are a few examples of U-Net adaptations. All of these
adaptations have some changes in common such as the ways in which
the path of information flow is altered and an encoder-decoder struc-
ture with skip connections. In this work, the author was inspired by
the success of ResNet and R2-UNet [5] employing modified residual
blocks. Here, the two convolutional layers in one block are sharing the
same weights [5].
For the segmentation of cell nuclei images [6], the intersection over
union (IoU) is used as a quality matrix for increasing the performance of
Unet, ResUNet, and DeepLab. The proposed model has a mean IoU value
of 0.9005 for validation dataset.
19.2 Methods
The Unet framework mainly inspired our architecture. We utilized
EfficientNet-B0 as an encoding part. Like Unet, encoder branch and
decoder branch features are concatenated. We chose EfficientNet as the
encoder because it proposes an effective compound scaling strategy that
effectively scales up a baseline ConvNet to any target resource limitations
in a more principled manner while retaining model efficiency.
The ImageNet dataset containing color images was used for training
the original EfficientNet. The images had a resolution of 224×224. In the
present study, a pre-trained EfficientNet model was implemented as an
encoder part of Unet to reduce the training time. The EfficientNet-B0 with
low parameter quantity was applied owing to limited complexity and an
easily identifiable nature of the neuronal images.
19.3 Dataset
The dataset consists of 606 grayscale images with an image size of 520 ×
704 pixels. We chose 364 photos at random for training, 61 for validation,
and 60 for testing. Image count of ‘shsy5y’, ‘Astro’, ‘Cort’ was 155, 131, and
320, respectively. The neuroblastoma cell line Cort consistently has the
lowest precision scores of the three cancer cell lines. This could be because
neuronal cells have a distinct, uneven, and concave look, making them dif-
ficult to segment using standard mask heads.
19.4 Implementation Details

Data augmentation is being performed which includes random cropping,
random horizontal/vertical flipping, 35° rotation, and further enhanced
accuracy normalization is used on the training images with various
schemes. In this study, the call back function was implemented, which
monetarized the validation loss for 150 training epochs. If there is not a
significant change in this value, then training is terminated. The training
and testing images have a resolution of 256 × 256 pixels. We used binary
cross-entropy (BCE) with sigmoid as a loss function and Adam as an
optimizer. The initial learning rate was set to 1e-4 to optimize the model
weights and PyTorch on Tesla P100-PCIE-16GB was used to implement
the model.
19.5 Evaluation Metrics

Average Precision (AP) [7] is used as a metric to evaluate both instances’
segmentation performance. For instance, in segmentation, we employ
APmask in Equation 19.1 to represent the AP at the mask and the IoU
between the ground truth mask and each predicted segmentation mask.
1 N
AIOUαmask= ∑i =1 IOUαmask (19.1)
N
19.6 Result
For this work of semantic segmentation, we used a variety of designs
and the results are given in Table 19.1. Our investigation began with the
DeepLabV3+ framework. For extraction of features, we initially utilized
EfficientNet-B0 as an encoder. We experimented with several encoder back-
bones such as EfficientNet-B0 [8], ResNet-50 [9], InceptionResNetV2 [10],
Table 19.1 Result comparison in terms of mean intersection over union (IoU),
F1-score, and accuracy on the neuronal dataset.
Model F1-score IoU Accuracy
Unet with EfficientNet-B0 as encoder 0.7286 0.573 0.9466
(binary cross entropy logits)
Unet with ResNet-50 as encoder (binary 0.7053 0.5447 0.9254
cross entropy logits)
Unet with Inception-ResNet-v2 as 0.7264 0.5704 0.9550
encoder (binary cross entropy logits)
Unet with EfficientNet-B0 as encoder 0.7655 0.6201 0.9493
(dice loss)
PSPNet with EfficientNet-B2 as encoder 0.6075 0.4362 0.9321
PSPNet with InceptionNet as encoder 0.6579 0.4902 0.9311
DeeplabV3+ with EfficientNet-B5 as 0.7089 0.5490 0.9351
encoder (binary cross entropy logits)
Unet with Inception-ResNet-v2 as 0.7283 0.5727 0.9166
encoder (dice loss)
DeeplabV3+ with EfficientNet-B5 as 0.7265 0.5705 0.9351
encoder (dice loss)
and EfficientNet-B5 with the renowned Unet decoder. From Table 19.1,
it can be shown that the Unet decoder outperforms the DeepLabV3+
decoder. That is because Unet now has additional low-level encoder char-
acteristics that are essential in analyzing complicated scenes with multiple
dense objects.
We further analyze the class-wise segmentation performance of the
best-performing model on the three classes. We obtain class-wise IoU on
the validation dataset, shown in Table 19.2. It can be observed that the Cort
Table 19.2 Class wise IoU on neuronal dataset with efficient Unet as model.
Neuronal cells shsy5y Astro Cort
Mean IoU 0.4580 0.5474 0.2948
cell has the lowest IoU. The Cort cell image contains different geometry
types compared to Astro and sphy5y. Besides the circular geometry of Cort
cells, thread-like structures are also present in the image, which is similar
to Astro and shyishy5 geometry. This makes model prediction clouded.
The model’s accuracy is further hampered by a small dataset. The ground
truth segmentation maps of a few images, as well as their anticipated seg-
mentation maps, are shown in Figure 19.2.
The first column displays the input photos for a single cell. The ground
truth and anticipated segmentation maps are shown in the second and
third columns, respectively. The cell type is in the order Astro, Cort,
Shsy5y from top to bottom.
Figure 19.2 Result of semantic segmentation on neuronal validation dataset with efficient
Unet as architecture.
19.7 Conclusion
This study employs multiple kinds of encoders for image segmentation
and is based on Unet architecture. It provides a comprehensive procedure
that includes selecting a data set, acquiring a training set, training a deep
convolutional neural network, and segmenting cell images using the con-
volution neural network. Finally, following cell image segmentation, the
Unet with various encoders was utilized to generate the resultant image,
and a basic analysis of the acquired values was performed. This technique
achieves high-precision semantic segmentation of shsy5y cells and Astro
pictures compared to traditional image segmentation.
References
1. S. Przedborski, M. Vila, and V. Jackson-Lewis, ‘Series Introduction:
Neurodegeneration: What is it and where are we?’, J. Clin. Invest., vol. 111,
no. 1, p. 3, Jan. 2003, doi: 10.1172/JCI17522.
2. W. H. Organization, ‘Neurological disorders: public health challenges’, World
Health Organization, 2006.
3. V. L. Feigin et al., ‘Global, regional, and national burden of neurological dis-
orders, 1990–2016: a systematic analysis for the Global Burden of Disease
Study 2016’, Lancet Neurol., vol. 18, no. 5, pp. 459–480, May 2019, doi:
10.1016/S1474-4422(18)30499-X.
4. Z. Zeng, W. Xie, Y. Zhang, and Y. Lu, ‘RIC-Unet: An Improved Neural
Network Based on Unet for Nuclei Segmentation in Histology Images’,
IEEE Access, vol. 7, pp. 21420–21428, 2019, doi: 10.1109/ACCESS.2019.
2896920.
5. J. Zhuang, ‘LadderNet: Multi-path networks based on U-Net for medical
image segmentation’, Oct. 2018, Accessed: Dec. 13, 2021. [Online]. Available:
https://arxiv.org/abs/1810.07810v4.
6. C. A. R. Goyzueta, J. E. C. De la Cruz, and W. A. M. Machaca, ‘Integration
of U-Net, ResU-Net and DeepLab Architectures with Intersection Over
Union metric for Cells Nuclei Image Segmentation’, pp. 1–4, Nov. 2021, doi:
10.1109/EIRCON52903.2021.9613150.
7. M. Everingham, L. Van Gool, C. K. I. Williams, J. Winn, and A. Zisserman,
‘The pascal visual object classes (VOC) challenge’, Int. J. Comput. Vis., vol. 88,
no. 2, pp. 303–338, Sep. 2010, doi: 10.1007/s11263-009-0275-4.
8. M. Tan and Q. V. Le, ‘EfficientNet: Rethinking Model Scaling for Convolutional
Neural Networks’, 36th Int. Conf. Mach. Learn. ICML 2019, vol. 2019-June,
pp. 10691–10700, May 2019, Accessed: Dec. 13, 2021. [Online]. Available:
https://arxiv.org/abs/1905.11946v5.
9. K. He, X. Zhang, S. Ren, and J. Sun, ‘Deep Residual Learning for Image
Recognition’, Proc. IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit.,
vol. 2016-December, pp. 770–778, Dec. 2015, doi: 10.1109/CVPR.2016.90.
10. C. Szegedy, S. Ioffe, V. Vanhoucke, and A. A. Alemi, ‘Inception-v4, Inception-
ResNet and the Impact of Residual Connections on Learning’, 31st AAAI
Conf. Artif. Intell. AAAI 2017, pp. 4278–4284, Feb. 2016, Accessed: Dec. 15,
2021. [Online]. Available: https://arxiv.org/abs/1602.07261v2.
20
Automorphisms of Some
Non-Abelian p−Groups of Order p4
Muniya Sansanwal1*, Harsha Arora2 and Mahender Singh3
Department of Mathematics, JJTU Jhunjhunu, Rajasthan, India
1
Department of Mathematics, Govt. College for Women Hisar, Haryana, India

2
3
Department of Mathematics, Om Sterling Global University, Hisar, Haryana, India
Abstract
This paper contains the number of automorphisms of several non-Abelian groups
of order p4, p-odd prime are computed and GAP (Group Algorithm Programming)
software has been used for the verification of a number of automorphisms.
Keywords: -p-groups, automorphism, semi-direct product
MSC subject classification 2000: 20D45, 20D60
20.1 Introduction
Let G be a p-group of order p4, p-odd prime and Aut(G) represent the group
of all automorphisms of a given group G. There are many research papers
in the literature related to the automorphisms of p-groups, for instance [1,
3, 4], etc. In [1], the automorphisms of groups of order p3 are computed
along with the automorphisms of abelian groups of order p4. The present
paper is an extension of the research work in [1]. In this paper, we will
compute the automorphisms of groups of order p4. William Burnside [2]
classified all groups having order p4. By using these classifications we derive
the automorphisms of some groups having order p4. We divide the derived
results into two parts. In these two parts, one part is dedicated to the cat-
egorization of all p−groups of order p4 and the other part is dedicated to
investigating the number of automorphisms of several non-Abelian groups
with order p4.
*Corresponding author: muniyarodhan15@gmail.com
291
20.2 Categorization of p-Groups with Order p4

According to the categorization derived by Burnside [2], if p is an odd
prime, then there will be 15 groups of order p4. Five of these are abelian
and the rest are non-abelian, as given below.
Abelian Groups:
4
G1 = < x1 : x1p = 1 > ≅ Z p4 .
3
G2 = < x1 , x2 : x1p= x=
p
2 1, x=
1x 2 x2 x1 > ≅ Z p3 × Z p .
2 2
p p
G3 = < x1 , x2 : x=
1 x=
1 1, x=
1x 2 x2 x1 > ≅ Z p2 × Z p2 .
2
G4 = < x1 , x2 , x3 : x1p= x= 2
p
x=p
3 1, x1x2 = x2 x1 , x2 x3
= x= 3 x 2 , x1x 3 x3 x1 > ≅ Z p2 × Z p × Z p .
p p p p
G5 = < x1 , x2 , x3 , x 4 : x=
1 x= 2 x=3 x=
4= 1, x1x2 x= 2 x1 , x1x 3 x3 x1 ,
x1x 4 = x= 4 x1 , x 2 x 3 x=
3 x2 , x2 x 4 x 4 x2 , x3 x 4 = x 4 x3
> ≅ Zp × Zp × Zp × Zp.
Non-Abelian Groups:
3 2
p
G6 =< x1 , x2 : x1= p
x=
2 , x2 x1 x11+ p x2 > ≅ Z p3 φ Z p ,
1=
φ ( y ) ↔ (1 + p2 ) y .
2 2
p
G7 =< x1 , x2 : x=
1
p
x=
2 , x2 x1 x11+ p x2 > ≅ Z p2 φ Z p2 ,
1=
φ ( y ) ↔ (1 + p) y .
2
p p
G8 =< x1 , x2 , x3 : x=
1 x=2 = x3p 1= , x1x2 x2 x1 , x2 x3 = x3 x2 ,
x3 x1 = x1x2 x3 > ≅ (Z p × Z p ) Z p2 .
p p p p
G9 =< x1 , x2 , x3 , x 4 : x=
1 x=2 x=3 x=
4 1, x 4 x3 = x3 x 4 , x2 x 4 = x 4 x2 ,
= x1x 4 x= 4 x1 , x 2 x 3 x3 x2 , x2 x1x3 = x3 x1 ,
x=1x 2 x2 x1 > ≅ ((Z Z p × Z p ) Z p ) × Z p .
p p p p
For pp >> 33,,G
G1010 =
<< xx1 ,,xx2 ,,xx3 ,,xx4 :: xx=
= 1 2 3 4 =
1p
1
xx=
2p
=2
xx=
3p
=3
x==
x=
=4 1,, xx44xx33 xx=
4p 1 2x
= 3 x 4 , x 4 x2
2 x3 x 4 , x 4 x2
xx1xx2 xx4 ,,
1 2 4
x3 x= 2 x3 x 4 , x 4 x2 x1x= 2 x 4 , x1x 4
= x1x 4 xx=
4x
= 1 , x2 x3
4 x1 , x 2 x 3
xx=
3x
= 2 , x1x 3
3 x 2 ,2 x1x 3
xx=
3x
= 1 , x1x 2
3 x1 , x1x 2
xx2 xx1 >>≅≅ ((ZZp ×× ZZp ×× ZZp )Z
2 1 p p p )Z
ZZp ,,
p
−p
> ≅ (Z p × Z p × Z p )ZZ p ,If p = 3 then < x1 , x2 x3 : x1p= x= 2
p
x =p
3 = 1, x x
1 2 x= x ,
2 1 1 3x x x 3 1 2 x 2 x 3 = x 3 x1 x 2 > .
x x ,
x=
2 x1 , x1x 3 x3 x1x2 , x2 x3 = x3 x1− p x2 > .
Automorphisms of Some Non-Abelian p−Groups of Order p4 293
2
1+ p
G11 = < x1 , x2 , x3 : x1p= x= p
2 x=
=
3
p
1, x2 x1 x= 1 x 2 , x1x 3 x3 x1 ,
x 2 x 3 = x 3 x 2 > ≅ ( Z p2  Z p ) × Z p .
2
G12 = < x1p= x=
p
2
p
x=
3=1, x2 x1 x=
1x 2 , x1x 3 x3 x1 ,
x3 x2 x1p x2 x3 > ≅ (Z p2 × Z p ) Z p .
=
2
1+ p
G13 = < x1 , x2 , x3 : x1p= x=
p
2
p
x=
=
3 1, x2 x1 x=
1 x 2 , x 3 x1 x1x2 x3 ,
x=
2 x3 x3 x2 > ≅ (Z p2 × Z p )φ1 Z p , φ1(z ) ↔ (1,1, 0)z .
2
1+ p
For p > 3,G14 = p
< x1 , x2 , x3 : x1= p
x=
2
p
x=
=
3 1, x2 x1 x=
1 x 2 , x 3 x1 x11+ p
, x3 x2 x1p x2 x3 > ≅ (Z p2  Z p )φ2 Z p , φ2 (z ) ↔ (1,1,1)z ,
x2 x3=
2
If p = 3 then < x1 , x2 , x3 : x1p = x2p =
1, x3p x=
= p
1 , x1x 2 x2 x11+ p ,
x1x3 = x3 x1x2−1 , x=
3 x2 x2 x3 > .
2
1+ p
Forp p> >3,3G,G
15 = < <x1x, x,2x, x,3x: x: 1xp=p=2
x= p
p x= = p
1, x2 x1 x=
p 1, x 2 x1 1+xp2 , x 3 x1 x11x+1dp+dp
x2xx3x, ,x3xx2x
15 =
2x= 3x== 1x=
1 2 3 1 2 3 1 x 2 , x 3 x1 1 2 3 3 2
1+ p 1+dp
x1 x2 , x3 x1 x1 x2 x3= , x3 x2 x1 dpdp
x x > ≅ ( Z  Z ) Z , φ ( z ) ↔ (1 ,1 , d ) z
where
z d ≢
= x1 2x22x2 > ≅ (Z p 2 Z φ3 Z p , φ3 (z ) ↔ (1,1,d )
2 p )φ 3 p 3
p
1,d ) z 0,1(mod p), 2
p
−p
If p= 3,< x1 , x2 , x3 ; x1p= x= p
2= 1, x3p x= 1 , x1 x 2 x2 x11+ p , x1x3 = x3 x1x2−1 ,x= 3 x2 x2 x3 > .
1+ p −1
1 , x1x 3 = x 3 x1x 2 ,x=3 x2 x2 x3 > .
20.3 Number of Automorphisms of Some

Non-Abelian Groups of Order p4
The automorphisms of five abelian groups, namely G1, G2, G3, G4, and G5, of
order p4 are already computed in the paper [1]. In this paper, we compute
the automorphisms of several non-abelian p−groups with order p4, namely
G9, G10, and G11 denoted above.
20.3.1 Computation of Automorphisms of Group G9
G9 = < x1 , x2 , x3 , x 4 ; x1p = x=2

p
x=3
p p
x=
4= 1, x 4 x3 x= 3 x 4 , x2 x 4 x 4 x2 , x1x 4
== x=
4 x1 , x 2 x 3 x3 x= 2 , x 2 x1x 3 x=3 x1 , x1x 2 x 2 x1 > ≅ ( ( Z p × Z p ) Z p ) × Z p .
We can deduce relations by using generator relations from the structure

description of G9 and elementary calculations:
x3i x1j = x1j x2(ij ) x3i ,

n(n−1)
( ij )
(x1i x3j )n = x1(ni ) x2 2 x (nj )
3 .
Now, we shall investigate the automorphism group ( (Z p × Z p ) Z p ) × Z p.

(
Let ψ ∈ Aut ( (Z p × Z p ) Z p ) × Z p be defined by: )
 x1 → x1i x2j x3k x 4l ; i, j, k, l ∈ Z p ,
 m n q r
 x2 → x1 x2 x3 x 4 ; m, n, q , r ∈ Z p ,
ψ : s t u v
 x3 → x1 x2 x3 x 4 ; s, t , u, v ∈ Z p ,
x → x w x x x y x z ; w, ,x , y , z ∈ Zp.
 4 1 2 3 4
As ψ is automorphism, ψ should preserve the relation x2x1x3 = x3x1, that is
ψ (x2 )ψ (x1 )ψ (x3 ) = ψ (x3 )ψ (x1 )

m n q r i j k l s t u v
⇒ x x x x x x x x x x x x = x1s x2t x3u x 4v x1i x2j x3k x 4l
1 2 3 4 1 2 3 4 1 2 3 4
⇒ x1(m+i +s ) x2(n+qi + j+qs +ks +t ) x3(q+k +u ) x 4(r +l +v ) = x1( s +i ) x2(t +ui + j ) x3(u+k ) x 4(v +l )
⇒ x1m x2(n+qi +qs +ks ) x3q x 4r = x2(ui )
⇒ m ≡ 0(mod p),
n + qi + qs + ks ≡ ui(mod p),
q ≡ 0(mod p) and
r ≡ 0(mod p)
⇒ n + ks ≡ ui(mod p)
or n ≡ iu − sk(mod p)
As n ≢ 0(mod p), because if n ≡ 0(mod p) then order of (ψ (x2 ) ) ≠ p.
⇒ n = iu - sk ≢ 0(mod p).
Also, x4x3 = x3x4 then ψ(x4)ψ(x3) = ψ(x3)ψ(x4)
⇒ x1w x2x x3y x 4z x1s x2t x3u x 4v =

x1s x2t x3u x 4v x1w x2x x3y x 4z
⇒ x1(w +s ) x2( x + ys +t ) x3( y +u ) x 4( z +v ) = x1( s +w ) x2(t +uw + x ) x3(u+ y ) x 4(v +z )
⇒ x2( ys ) = x2(uw )
⇒ ys ≡ uw(mod p)
⇒ ys − uw ≡ 0(mod p).
Further, x1x4 = x4x1 implies ψ(x1)ψ(x4) = ψ(x4)ψ(x1)
⇒ x1i x2j x3k x 4l x1w x2x x3y x 4z =

x1w x2x x3y x 4z x1i x2j x3k x 4l
⇒ x1(i +w ) x2( j+kw + x ) x3(k + y ) x 4(l +z ) = x1(w +i ) x2( x + yi + j ) x3( y +k ) x 4( z +l )
⇒ x2(kw ) = x2( yi )
⇒ kw ≡ yi(mod p)
⇒ yi − kw ≡ 0(mod p).
Now, ys − uw ≡ 0(mod p) and yi − kw ≡ 0(mod p) are two homogeneous

equations with determinant iu − sk ≠ 0. Hence, these equations have only
trivial solutions. ⇒ y = w ≡ 0(mod p).
Next, we check the kernel of ψ. Let x1h1 x2h2 x3h3 x 4h4 be any element of
((Z p × Z p ) Z p ) × Z p so that it is mapped to identity.
ψ (x1h1 x1h2 x1h3 x1h4 ) = 1
⇒ ψ (x1 )h1ψ (x2 )h2ψ (x3 )h3ψ (x 4 )h4 =
1
⇒ (x1i x2j x3k x 4l )h1 (x1m x2n x3q x 4r )h2 (x1s x2t x3u x 4v )h3 (x1w x2x x3y x 4z )h4 = 1
h (h −1) h (h −1)
( jh1 + 1 1 ik +nh2 +th3 + 3 3 su + xh4 +kh1sh3 )
⇒ x1(ih1 +sh3 ) x2 2 2 x ( kh1 +uh3 ) (lh1 +vh3 + zh4 )
3 4 x =1
⇒ ih1 + sh3 ≡ 0(mod p), (20.1)
h1(h1 − 1) h (h − 1)
jh1 + ik + nh2 + th3 + 3 3 su + xh4 + kh1sh3 ≡ 0(mod p),
2 2
(20.2)
kh1 + uh3 ≡ 0(mod p), (20.3)
lh1 + vh3 + zh4 ≡ 0(mod p). (20.4)
Equations (20.1) and (20.3) are the system of two homogeneous equa-
tions with iu − sk ≠ 0 determinant, which gives only a trivial solution.
Hence, h1 = h3 = 0. From (20.4), we get
zh4 ≡ 0(mod p)
As
z ≢ 0(mod p),
we get
h4 = 0.
Further, from (20.2) we get nh2 ≡ 0(mod p), but n ≢ 0(mod p) ⇒ h2 = 0.
( )
So, if h ∈ ( (Z p × Z p ) Z p ) × Z p and ψ(h) = 1, it must be that h = 1.
Hence, the kernel is trivial and ψ is an automorphism with the con-
straints on the parameters already deduced. Now it is very easy to calculate
the order of the automorphism group. We have p choices for all j, l, t, v, x,
and p − 1 choices for z and i, u, s, and k will in some sense be equivalent
to a matrix in GL(Fp), which gives (p2 −p)(p2 − 1) choices for the elements.
( )
Therefore, we have that ψ ∈ Aut ( (Z p × Z p ) Z p ) × Z p is really defined as:
 x1 → x1i x2j x3k x 4l ; j, l ∈ Z p ,

 iu −sk
 x2 → x2 ,
ψ : s t u v
 x3 → x1 x2 x3 x 4 ; t , v ∈ Zp ,
x → x x x z ; x , z ∈ Z p , p  z.
 4 2 4
Aut ( ( (Z p × Z p ) Z p ) × Z=
p) p6 ( p − 1)2 ( p2 − 1).
20.3.2 Computation of Automorphisms of Group G10

p p p p
G10 = < x1 , x2 , x3 , x 4 ; x=
1 x=2 x=
3 x= 4 1, x 4 x3 = x2 x3 x 4 , x 4 x2
= x1x=
2 x 4 , x1x 4 x=
4 x1 , x 2 x 3 x=
3 x 2 , x1x 3 x=3 x1 , x1x 2 x2 x1
> ≅ (Z p × Z p × Z p ) Z p
By using our relations from the structure of G10 and elementary calcula-
tions, we can deduce some useful relations:
x 4i x2j = x1(ij ) x2j x 4i ,

ij (i −1)
( )
i j (ij ) j i
x x =x
4 3 1
2
2 x xx ,
3 4
 n(n−1) n(n−1)l (l −1) n(n−1)l (n−2) 2   n(n−1) 
i j k l n  ni + jl + k+ kl   jn+ kl 
 ( kn) ( jn)
(x x x x ) = x
1 2 3 4

1
2 4 6 
x 
2
2
x 3 x 4 .
We begin our study of automorphism group (Zp × Zp × Zp) ⋊ Zp by using

the above relations.
Let ψ ∈ Aut ( (Z p × Z p × Z p ) Z p ) be defined by:
 x1 → x1i x2j x3k x 4l ; i, j, k, l ∈ Z p ,

 m n q r
 x2 → x1 x2 x3 x 4 ; m, n, q , r ∈ Z p ,
ψ : s t u v
 x3 → x1 x2 x3 x 4 ; s, t , u, v ∈ Z p ,
x → x w x x x y x z ; w, x, y, z ∈ Zp.
 4 1 2 3 4
Since ψ is an automorphism, ψ should preserve the relation x4x3 = x2x3x4,

hence
ψ(x4)ψ(x3) = ψ(x2)ψ(x3)ψ(x4)
ψ (x 4 )ψ (x3 ) = ψ (x2 )ψ (x3 )ψ (x 4 )
w x y z s t u v
x1m x2n x3q x 4r x1s x2t x3u x 4v x1w x2x x3y x 4z
⇒x x x x x x x x =
1 2 3 4 1 2 3 4
zu ( z −1)
( zt + )
( zu )
⇒ x1 2 x 2
ru (r −1) vy (v −1) ry (r −1)
(m+rt +vx + + +rx +rvy + )
(n+ru +vy +ry ) q r
=x 1
2 2 x x x 2
2 3 4
zu(z − 1) ru(r − 1) vy(v − 1) ry(r − 1)

⇒ zt + ≡ m + rt + vx + + + rx + rvy + (mod p),
2 2 2 2
(20.5)
zu ≡ n + ru + vy + ry (mod p), (20.6)
q ≡ 0(mod p) and (20.7)
r ≡ 0 (mod p) (20.8)
using (20.7) and (20.8) in (20.5) and (20.6), we get
zu(z − 1) vy(v − 1)
⇒ zt + ≡ m + vx + (mod p), (20.9)
2 2
zu ≡ n + vy(mod p) (20.10)
Also,
x 4 x2 = x1x2 x 4
⇒ ψ (x 4 )ψ (x2 ) = ψ (x1 )ψ (x2 )ψ (x 4 )
⇒ x1w x2x x3y x 4z x1m x2n x3q x 4r = x1i x2j x3k x 4l x1m x2n x3q x 4r x1w x2x x3y x 4z
 ly (l −1) 
 i +m+w +ln+lx + 
(w +m+ zn) ( x +n) y z 2  ( j +n+ x +ly ) ( k + y ) (l + z )
⇒x 1 x
2 x x =x
3 4

1 x 2 x
3 x4
 ly (l −1) 
 i +ln+lx + 
2  ( j +ly ) k l
⇒ x1( zn) =
x 
1 x 2 x x
3 4
ly(l − 1)
⇒ zn ≡ i + ln + lx + (mod p), (20.11)
2
j + ly ≡ 0(mod p), (20.12)
k ≡ 0(mod p) and (20.13)
l ≡ 0(mod p) (20.14)
Using Equations (20.13) and (20.14) in (20.11) and (20.12), we get
⇒ zn ≡ i(mod p) or i ≡ zn(mod p) and (20.15)
j ≡ 0(mod p). (20.16)
Since j, k, l ≡ 0(mod p), therefore i ≢ 0(mod p) because if i ≡ 0(mod p),

then Ord(x1) ≠ p. Further, from (20.15), i = zn ≢ 0(mod p), we have z ≢
0(mod p) and n ≢ 0(mod p).
Next, we have x2x3 = x3x2
⇒ ψ (x2 )ψ (x3 ) =
ψ (x3 )ψ (x2 )
⇒ x1m x2n x3q x 4r x1s x2t x3u x 4v =
x1s x2t x3u x 4v x1m x2n x3q x 4r
⇒ x1(m+s ) x2(n+t ) x3u x 4v =
x1( s +m+vn) x2(t +n) x3u x 4v
⇒ vn ≡ 0(mod p).
Now, n ≢ 0(mod p) gives v ≡ 0(mod p)

From (20.10), we can see zu ≡ n + vy(mod p), therefore zu ≡ n(mod p)
or n ≡ zu(mod p).
Also, from (20.15) i ≡ zn(mod p), we get i ≡ zzu(mod p). Thus, we have
I ≡ z2u(mod p).
Further, as n ≡ zu(mod p) and n ≢ 0(mod p), therefore u ≢ 0 (mod p).

zu(z − 1)
Substituting the value of v in (20.9), we get zt + ≡ m(modp) ≡
zu(z − 1) 2
m(mod p) or m ≡ zt + . Now, we shall find the constraints on the
2
parameters for which ψ has a trivial kernel. Let x1h1 x2h2 x3h3 x 4h4 ∈ ker (ψ ).
⇒ ψ (x1h1 x2h2 x3h3 x 4h4 ) =

1
⇒ ψ (x1 )h1ψ (x2 )h2ψ (x3 )h3ψ (x 4 )h4 =
1
⇒ (x1i x2j x3k x 4l )h1 (x1m x2n x3q x 4r )h2 (x1s x2t x3u x 4v )h3 (x1w x2x x3y x 4z )h4 = 1
h (h −1) h (h −1) z ( z −1) h4 (h4 −1)(h4 −2) 2
(ih1 +mh2 + sh3 +wh4 + 4 4 xz + 4 4 y+ yz )
⇒ x1 2 4 6
h (h −1)
(nh2 +th3 + xh4 + 4 4 yz )
(uh3 + yh4 ) (zzh4 )
x 2
2 x 3 x4 =1
h4 (h4 − 1) h (h − 1)z (z − 1)
⇒ ih1 + mh2 + sh3 + wh4 + xz + 4 4
2 4
h (h − 1)(h4 − 2) 2
y+ 4 4 yz ≡ 0(mod p), (20.17)
6
h4 (h4 − 1)
nh2 + th3 + xh4 + yz ≡ 0(mod p), (20.18)
2
uh3 + yh4 ≡ 0(mod p), (20.19)
zh4 ≡ 0(mod p). (20.20)
From (20.20), we have h4 ≡ 0(mod p) for z ≢ 0(mod p) and from (20.19),

we get uh3 ≡ 0(mod p). Since u ≢ 0(mod p), therefore h3 = 0. Further, from
(20.18), we have nh2 ≡ 0(mod p). As n ≢ 0(mod p), we have h2 = 0. From
(20.17), ih1 ≡ 0(mod p), but i ≢ 0(mod p), thus h1 = 0.
So, if h ∈ ( (Z p × Z p × Z p ) Z p ) and ψ(h) = 1, it must be that h = 1, hence
the kernel is trivial and ψ is an automorphism with the constraints we have
already deduced. Thus, for ψ to be an automorphism there are p choices for
all s, t, w, x, y, and p − 1 choices for both u and z. Therefore, we have that
ψ ∈ Aut ((Z p × Z p × Z p ) Z p ) is really defined as:
 x1 → x1nz ; nz  0(mod p),

 zu ( z −1)
 zt +
ψ :  x2 → x1
z x2zu ,
 x3 → x1s x2t x3u ; s, t , u ∈ Z p , p  u,
 w x y z
 x 4 → x1 x2 x3 x 4 ; w , x , y , z ∈ Z p , p  z.
Aut ( (Z p × Z p × Z p ) Z=
p) p5 ( p − 1)2 .
20.3.3 Computation of Automorphisms of G11

2
1+ p
G11 = < x1 , x2 , x3 ; x1p= x= 2
p
x==
3
p
1, x2 x1 x= 1 x 2 , x1x 3 x3 x1 , x2 x3
= x 3 x 2 > ≅ ( Z p2  Z p ) × Z p
We can deduce some useful relations by using the generator relations of

G11 and elementary calculations:
x2j x1i = x1(i + jip) x2j ,

n(n−1)
(ni + ijp )
(x1i x2j )n = x1 2 x(nj )
2 .
With these relations, we begin our study of the automorphisms of

( Z p2  Z p ) × Z p .
( )
Let ψ ∈ Aut (Z p2  Z p ) × Z p , be defined by:
 x1 → x1i x2j x3k ; i ∈ Z p2 , j , k ∈ Z p ,


ψ :  x2 → x1l x2m x3n ; l ∈ Z p2 , m , n ∈ Z p ,
 q r s
 x3 → x1 x2 x3 ; q ∈ Z p2 , r , s ∈ Z p .
Now, ord(x1) = ord(ψ(x1)) = p2.

Therefore, (x1i x2j x3k ) p ≠ 1. Hence
x1pi ≠ 1
⇒ p  i.
Also, the order of x2 ∈ (Z p2  Z p ) × Z p is p, so the order of ψ(x2) is

also p.
⇒ (x1l x2m x3n ) p =

1
⇒ x1lp = 1
⇒ p/l
⇒ l ⇒ pt ; for some t ∈ Z p .
Order of x3 is also p, so Order of ψ(x3) is p
⇒ (x1q x2r x3s ) p =

1
⇒ x1qp = 1
⇒ p/q
⇒q=pu; for some u ∈ Z p .
As ψ is an automorphism, ψ should preserve the relation x2 x1 = x11+ p x2 ,

thus we get
ψ (x2 )ψ (x1 ) = ψ (x1 )1+ pψ (x2 )

⇒ x1l x2m x3n x1i x2j x3k =
(x1i x2j x3k )1+ p x1l x2m x3n
⇒ x1(l +i +mip) x2(m+ j ) x3(n+k ) =
x1( pi +i +l ) x2( j+m) x3(k +n)
⇒ x1(mip) =
x1( pi )
⇒ imp ≡ pi(mod p2 )
⇒ im ≡ i(mod p)
Since p  i ⇒ i ≡ 0(mod p)
⇒ m ≡ 1(mod p).
And x3 x1 = x1x3 , then
ψ (x3 )ψ (x1 ) = ψ (x1 )ψ (x3 )
⇒ x1q x2r x3s x1i x2j x3k = x1i x2j x3k x1q x2r x3s
⇒ x1(q+i +rip) x2(r + j ) x3( s +k ) = x1(i +q+ jqp) x2( j+r ) x3(k +s )
⇒ x1(rip) =
x1( jqp)
x1( jpup)
⇒ x1(rip) =
⇒ x1(rip) =
1
⇒ p/ri
but p  i ⇒ p/r
⇒ r ≡ 0(mod p).
Next, we find the constraints for parameters using the kernel. Let
x1x x2y x3z be any element of (Z p2  Z p ) × Z p that it is mapped to identity.
⇒ ψ (x1x x2y x3z ) =

1
⇒ ψ (x1 )xψ (x2 ) yψ (x3 )z =
1
⇒ (x1i x2j x3k )x (x1( pt ) x2m x3n ) y (x1( pu ) x3s )z = 1
x ( x −1)
( xi + ijp+ pty + puz )
( xj +my ) ( xk +ny + sz )
⇒ x1 2 x 2 x 3 =1
x(x − 1)
⇒ xi + ijp + pty + puz = 0(mod p2 ), (20.21)
2
xj + my = 0(mod p), (20.22)
xk + ny + sz = 0(mod p). (20.23)
From (20.21), we get
x(x − 1)
p2 /xi + ijp + pty + puz
2
x(x − 1)
⇒ p/p2 /xi + ijp + pty + puz
2
x(x − 1)
⇒ p/xi + ijp + pty + puz
2
⇒ p/xi
But p  i
⇒ p/x
⇒x= 0(mod p).
Next, from (20.22) we have
my ≡ 0(mod p)
But p f m
⇒ y ≡ 0(mod p).
Also, from (20.23) we obtain sz ≡ 0(mod p), but p f s, therefore z ≡

0(mod p).
So, if h ∈ (Z p2  Z p ) × Z p and ψ(h) = 1, it must be that h = 1, hence the
kernel is trivial and ψ is an automorphism with the restrictions we have
already deduced on the parameters. Now, we have here p2 − p choices for i
and p choices for all l, q, j, k, n, and p − 1 choices for s. Therefore, we have
that ψ ∈ Aut (Z p2  Z p ) × Z p is really defined as:
 x1 → x1i x2j x3k ; p  i , j, k ∈ Z p ,


ψ :  x2 → x1pt x2 x3n ; t, n ∈ Zp ,
 pu s
 x3 → x1 x3 ; u ∈ Z p , p  s.
p6 ( p − 1)2 .
|Aut (Z p2  Z p ) × Z p | =
References
1. H. Arora and R. Karan: What is the probability an automorphism fixes a group
element?, Communications in Algebra, 45(3), 1141–1150 (2017).
2. William Burnside: Theory of groups of finite order, Cambridge University
Press, first edition, 1897. Reprinted 2010 through Nabu Press.
3. Geir T. Helleloid Automorphism Groups of Finite p-Groups: Structure and
Applications, arxiv: 0711.2816 (2007).
4. Hans Liebeck: The Automorphism Group of finite p-groups, Journal of Algebra,
4, 426–432 (1966).
21
Viscoelastic Equation of
p-Laplacian Hyperbolic Type with
Logarithmic Source Term
Nazlı Irkıl* and Erhan Pişkin†
Department of Mathematics, Dicle University, Diyarbakır, Turkey
Abstract
This paper aims to address the problem with viscolelastic p-Laplacian type equa-
tions with logarithmic nonlinearity,
t
u tt +
∫
0
g(t − s)∆ u(s) ds − div((1 + ∇ p − 2 )∇u) − ∆u t = |u|p − 2 uln|u|
in which p ≥ 2, under a convenient hypotheses on g (t). Under suitable conditions,

we discuss global existence and blow up the results.
Keywords: Existence, blow up, viscoelastic equation, p–Laplacian type equation,
logarithmic nonlinearity
21.1 Introduction
Let Ω be a bounded domain in Rn (for n ≥ 1) with smooth boundary Σ =
∂Ω. We deal with the following problem in the initial boundary problem
for (x, t) ∈ Ω × R+:
*Corresponding author: nazliirkil@gmail.com

†
Corresponding author: episkin@dicle.edu.tr
305
 t
 tt
0
∫
 u + g(t − s) ∆ u(s) ds − div((1 + ∇( p − 2) )∇u) − ∆u = |u|( p − 2)uln|u|,
t

 u(x,0) = u 0 (x), u t (x , 0) = u1 (x),

 u(x , t) = 0, x ∈∑ × R + ,
The memory kernel g(t) is a real function which has typical properties.
In (21.1), if we take ln |u| = 1 and replace the second order as the fourth
order, we have
t
2
∫
utt + ∆ u + g(t − s) ∆ u(s) ds − div(∇ p − 2∇u) − ∆u t = f (u).
0
Equation (21.2) was named as a fourth order weak viscoelastic plate equa-
tion with a lower order perturbation of p-Laplacian type and it can be taken
as a general form of the one-dimensional nonlinear equation of elastoplastic
microstructure flows. This type of equation was first considered by Andrade
et al. [3]. They deal with interplay of the p-Laplacian operator with viscoelastic
terms. The authors obtained existence and decay results of solutions with gt(t)
≤ cg (t) . Later results about qualitative theory results of solutions to (21.2) type
equations were obtained by different authors (see [4, 13, 14, 18, 19, 26]).
The logarithmic wave equations which were introduced by Bialynicki-
Birula and Mycielski in [5] are related with many areas of physics.
Logarithmic source term occurs naturally in inflation cosmology and
quantum mechanics, supersymmetric field theories, and nuclear physics
[7, 17]. Later, by the motivation of these works, analysis of logarithmic
source terms attracted many authors (see [8−10, 16, 23, 25, 32, 35]). There
is some work about mathematical behavior in viscoelasticity with logarith-
mic source terms. The references were given in our work for interested
readers (see [2, 6, 11, 12, 20, 31]).
The model (21.1) for g = 0 and without ∆u was studied, but there is
not a substantial number of work related with p-Laplacian hyperbolic type
equations. For contact problems on p-Laplacian type equations with loga-
rithmic source terms see ([15, 24, 30, 33, 34]).
The goal of our work is to prove the global existence of a weak solution
to the problem (21.1). Blow up results were established for suitable condi-
tions at E (0) < d. As far as know, there is no work which is related on p-
Laplacian hyperbolic type equations with viscoelastic terms and logarith-
mic source terms.
Viscoelastic Equation of p-Laplacian Hyperbolic Type 307
The structure of the work is as follows: to understand the expression in

the work more easily, firstly we give some definitions, notations, energy
functions, and some lemmas which will be used in our proof in Section
21.1. In part 2 and part 3, global existence results and blow up results were
stated for the solution of problem (21.1), respectively.
21.2 Preliminaries
In this part, we consider the notation and some well-known lemmas. We
define the inner product as
(v , u) =
∫ v(x )u(x )dx.
in Standard Lebesque Space L2 (Ω) and we denote:
(∇v, ∇u) = ∫ ∇v(x)∇u(x)dx
in H 01 (Ω ). The norm in L2 (Ω) and Lp (Ω) will be defined by ǁ.ǁ and ǁ.ǁ p. And
C will be various positive constants.
Lemma 21.1. [1, 22] Let q be a positive number, which implies that
2 ≤ q < ∞, for n = 1,2
2n
2≤q≤ , for n ≥ 3.
n−2
So, for ∀ u ∈ H 01 (Ω ), we have
ǁuǁq ≤ Cq ǁuǁ (21.3)
where Cq is a positive constant.

Now, we will consider some assumptions based on the kernel function g(t):
(A1) In accordance with g: [0, ∞) → R+ (g(t) is a nonnegative function),
we suppose that the function g ∈ C1 [0, ∞) and
∞
g (0) ≥ 0 ,
∫ g (s)ds = 1 − l > 0,
0
(21.4)
(A2) Let ϑ be a positive number which satisfies for t ≥ 0
g′ (t) ≤ ϑg(t). (21.5)

E(t) will be denoted total energy functional for the model investigated
in (21.1). It follows that
t
1 1 1  1
E(t ) = ||ut ||2 + ||∇ u||pp +  1 −
2 p 2  ∫0
g (s )ds  ||∇ u||2 + 2 ||u||pp
 p
1 1
+ ( g °∇ u)(t ) −
2 p ∫ u ln|u| dx
Ω
p (21.6)
We introduce the J(u) as a potential energy functional so that
t
1 1  1
p
p
J (u) = ||∇ u||p +  1 −
2  ∫0
g (s )ds  ||∇ u||2 + 2 ||u||pp
 p
1 1
+ ( g °∇ u)(t ) −
2 p ∫ u ln|u| dx.
Ω
p
(21.7)
We obtain for u ∈W01, p (Ω ),
t
 
p
I (u) = ||∇ u|| +  1 −
p
 ∫
0
g (s )ds  ||∇ u||2 + ( g °∇ u)(t ) −
 ∫ u ln|u| dx ,
Ω
p
(21.8)
where
t
( g °∇ u)(t ) = ∫ 0 g (t − s )||∇ u(s ) −∇ u(t )||2 ds (21.9)
Then, it is clearly concluded that for u ∈W01,p (Ω ),
t
1 1   I (u) 1
J (u) =  −   1 −
 2 p   ∫
0
g (s ) ||∇ u||2 +
 p
+ 2 ||u||pp
p
1 1
+  −  ( g °∇ u)(t ), (21.10)
 2 p
1
E(t ) = ||ut ||2 + J (u). (21.11)
2
According to potential well theory, the set stability for the problem
(21.1) was considered as:
U = {u ∈W01,p (Ω )| J (u) < d,I (u) > 0} ∪ {0}, (21.12)
So, we denote outer space of the potential well by
V = {u ∈W01,p (Ω )| J (u) < d,I (u) < 0}. (21.13)
The potential well depth can be considered as:

d= inf
1, p
sup(λu). (21.14)
u∈W0 ( Ω )\{0} λ >0
Similar to the result in [29], one has:
0 < d = inf J (u),

u∈N
Where according to the potential functional, well-known Nehari

Manifold N is given by:
N = {u ∈W01,p (Ω )\ {0}: I (u) = 0}.
Now, we state some definitions and lemmas. Lemma 21.5 and Lemma
21.6 are related with potential well theory results.
Lemma 21.2. [21]. Suppose that u is a function which satisfies u ∈W01,p (Ω )

and a is defined as positive number. Then,
( p − 2)a 2 a2 n
∫
Ω
u p ln |u|dx <
4π
||u||pp +
2π
||∇ u||pp − (1 + ln a)||u||pp + ||u||pp ln ||u||p
p
Definition 21.3. Let T > 0 and u be defined as the solution of problem

(21.1). If u holds
u ∈C([0 ,T ];W01,p (Ω )),ut ∈C([0 ,T ]; H 01 (Ω ),utt ∈([0 ,T ];W0−1,p ¢ (Ω )),

and
t
∫Ω
utt vdx +
∫Ω
(1 +∇ p−2 )∇ u∇ vdx −
∫ ∫ ∫ ∇ u∇ vdsdx + ∫ ∇ u∇ vdx = ∫ u
Ω 0 Ω Ω
t
Ω
p−2
uln |u|vdx,
for each test functions v ∈W01,p (Ω ), then u will be taken as a weak solution
for the (21.1).
Lemma 21.4. Assume that (A1) and (A2) conditions, which are related with
g(t) functional, hold. Then, the total energy of problem (21.1) decreases
over respect to t because
1 1 g (t )
E ′(t ) = ( g ′ °∇ u)(t ) − ||∇ ut ||2 − ||∇ u(t )||2
2 2 2
1 (21.15)
≤ − ||∇ ut ||2 ≤ 0.
2
where
( g ′ °∇ u)(t ) =
∫ g ′(t − s)||∇ u(s) −∇ u(t )|| dt
0
2
(21.16)
Proof. First, we multiply both sides of the equation by ut in the (21.1)

equation. Later, by integrating over t it yields that
t t
1 1
E(t ) = E(0) − ||∇ ut ||2 −
2 ∫ 0
2
g (t )||∇ u(t )|| +
2 ∫ ( g ′ °∇ u)(t )
0
1
≤ ||∇ ut ||2 (21.17)
2
Here, by direct calculations we obtained (21.16).

1 ,p
Lemma 21.5. For any u ∈W0 (Ω )\{0}, and ||u||pp ≠ 0, it gives
i) lim J (λu) = 0 , lim J (λu) = −∞ ,

λ →0 λ →∞
ii) For 0 < λ < ∞ we find a unique λ* such that
dJ (λu)
=0
d λ λ = λ*
iii) J (λu) functional is strictly decreases on λ* < λ < ∞, increases on
0 < λ < λ* and attains the maximum at λ = λ* and I (λu) > 0 for
0 < λ < λ*, I (λu) < 0 for λ* < λ < ∞ and the Nehari functional
dJ (λu)
I(λ*u) = 0, where λ = I (λu).
dλ
Proof. i) Thanks to the definition of J(u), we conclude that
t
1 1 
J ( λu ) =
p ∫
Ω
p
(λ∇ u) dx +  1 −
2  ∫
0
g (s )ds 
 ∫ (λ∇ u) dx
Ω
2
1 1 1
−
p ∫ (λu) ln |λu|dx + p ∫ (λu) dx + 2 λ ( g °∇ (λu))(t )
Ω
p
2
Ω
p 2
t
λ2   λ2 1
∫
p
= 1 − g (s )ds  || u||2 + ( g °∇ u)(t ) + 2 λ p ||u||p
2   2 p
0
p
λ λp 1 p
∫ ∫ (u) dx.
p
+ ||∇ u||p − (u) p ln|u| dx − λ lnλ p
(21.18)
p p p
Ω Ω
p
It is easy to get from ||u||p ≠ 0 that is lim J (λu) = 0 , lim J (λu) = −∞.
λ →0 λ →∞
ii) Now, differentiating J (λu) with respect to λ, it yields that
 t

dJ (λu)
dλ
= λ  1 −


∫
0
g (s )ds  || u||2 +( g °∇ u)(t ) + λ p − 2 ||∇ u||pp

−λ p−2
∫ (u) ln|u|dx − λ
Ω
p p−2
lnλ||∇ u||pp  .


(21.19)

dJ (λu)
Let λ −1 = φ (λu), so we obtain
dλ
dφ (λu)
dλ
= − λ p −3[(2 − p)||∇ u||pp +( p − 2)
∫ (u) ln|u|dx
Ω
p
p p
+ ( p − 2)lnλ||u|| + ||u|| ].
p p
dφ (λu)
Let = 0 , which implies
dλ
 (2 − p)||∇ u||pp +( p − 2) ∫ Ω (u) p ln|u| dx + ||u||pp 

λ1 = 0 and λ2 = exp  
 (2 − p)||u||pp
An elementary calculation shows that
dφ (λu) dφ (λu)
> 0 on (0 , λ2 ); < 0 on (λ2 , ∞).
dλ dλ
We observe from 2 ≤ p that lim φ (λu) = −∞. so that there is exactly

λ →∞
dJ (λu)
ϕ(λ*u) = 0, which means = 0.
d λ λ = λ*
iii) A simple corollary of the (ii) directly follows from
dJ (λu)
I ( λu ) = λ . (21.20)
dλ
Lemma 21.6. Assume that u ∈W01,p (Ω )\{0}, and ||u||pp ≠ 0. Then, we obtain
sup λu = d where
λ >0
n 2(n + p( l +1))− p 2 ( l +1)
d = [2π (l + 1)] e
2 2 .
Proof. If I(λ*u) = 0 and ||u||pp ≠ 0, from Lemma 21.4 and Equation (21.10),
we obtain
t
I (λ *u) ( p − 2)  
sup λu = J (λ *u) =
λ >0 p
+
2 p 
1 − ∫0
g (s )ds  ||∇λ *u)||2

1 p 1 1
+ 2 ||λ *u )||p +  −  ( g °∇ (λ *u))(t )
p  2 p
( p − 2) 1
≥ l ||∇λ *u)||2 + 2 ||λ *u)||pp
2p p
1 p
≥( 2 ||λ *u )||p . (21.21)
p
2
Since 0 < < 1, now by using the following inequality
p
xv ≤ x + 1
we have
2
||∇ u|| ≤ (||∇ u|| ) ≤ ||∇ u)||pp +1

2 p p
p (21.22)
By using Lp Logarithmic Sobolev inequality and definition of I(u) and

(21.22), we get
 t

I (u) ||∇ u)|| +  1 −
=


p
p
0
∫ 
 Ω
∫
g (s)ds  ||∇ u||2 + ( g °∇ u)(t ) − (u) p ln|u| dx
 ( p − 2)a 
2
a2
≥ ||∇ u)||pp + l ||∇ u||2 + ( g °∇ u)(t ) −   ||u )| | p
p − ||(∇ u)||pp
 4π  2π
n a2
+ (1 + ln a)||u||pp − ||u||pp ln ||u||p ≥ (l + 1 − (||∇ u||2 −1)
p 2π
n  ( p − 2)a2   p
+  (1 + ln a) −   − ln ||u||p  ||u||p (21.23)
p  4π  

Taking a = 2π (l + 1) , we find
n 1 (l + 1)( p − 2) 
I (u) ≥   1 + ln2π (l + 1) −
p
− ln ||u||p  ||u||p . (21.24)
p  2  2 
From I(λ*u) = 0 and (21.24), we conclude that
n 2(n + p ( l +1))− p 2 ( l +1))

ln[2π (l + 1)] e
2 2 − ln ||λ *u||p ≤ 0,
which satisfies that
n 2(n + p ( l +1))− p 2 ( l +1))

p
||λ *u|| ≥ [2π (l + 1)] e
p
2 2 = h. (21.25)
Thus, by combining (21.21) and (21.25), we obtain
n 2(n + p ( l +1))− p 2 ( l +1))

1 h
d > 2 [2π (l + 1)]2 e 2 = .
p p2
21.3 Global Existence Result

This part is related to global existence results. We established that the solu-
tions will exist globally for problem (21.1) for the case 0 < E(0) < d by con-
sidering the following lemma to state proof of Theorem 21.8.
Lemma 21.7. Assume that u0 ∈ U, u1 ∈ H 01 (Ω ). Let initial energy holds 0 <

E(0) < d. Moreover, u ∈ U where u is taken as a weak solution of Equation
(21.1).
Proof. For 0 < E(0) < d, we claim that u was taken as a weak solution for
Equation (21.1). Now, we will use the contradiction method. We suppose
that u ∉ U and t* are the smallest time of u (t*) ∉ U. By continuity of u(t),
we know clearly that u(t*) ∈ ∂U and u ∈ U for t ∈ [0, t*). By virtue of the
definition of U and the continuity of I(u) and J(u) about time, we have
either J(u(t*)) = d or I(u(t*)) = 0. Thanks to Lemma 21.3, it follows that
d > E(0) ≥ E (u(t *)) ≥ J(u(t *)) . (21.26)
Therefore, it is clear that the former case is not possible. Suppose that
I(u(t*)) = 0, then u (t*) ∈ N. By the definition of d, J(u (t*)) ≥ d was
obtained, which is in contradiction with J(u (t*)) < E(0) < d. So, we showed
that the other case is not possible as well.
Theorem 21.8. Let u0 ∈ U, u1 ∈ H 01 (Ω ). Then, Equation (21.1) with
initial boundary conditions has a global weak solution u(t ) ∈ L∞ (0 , ∞;W01,p (Ω )) with utt ∈
u(t ) ∈ L (0 , ∞;W01,p (Ω )) with utt ∈ L∞ (0 , ∞; H 01 ((Ω )) under the condition 0 < E (0) < d.
∞
Proof. Our purpose is to prove that ||∇ u||pp +l ||∇ u||2 + ||∇ u||pp is bounded
for not depending on t. By using Lemma 21.4, (21.11), definition of E(t),
and the condition 0 < E(0) < d, we obtain
1
E(t ) = ||ut ||2 + J(u(t )) ≤ E(t ) < d . (21.27)
2
According to conditions in Theorem 21.8, by using Lemma 21.7, for t ≥ 0

we obtain
u(t) ∈ U. (21.28)
We conclude from (21.10) and (21.28) that
t
I (u)  1 1   
d > J (u) = + − 1 −
p  2 p   ∫0
g (s )ds  ||∇ u||2

1 p 1 1
+ 2 ||u||p +  −  ( g °∇ u)(t )
p  2 p
( p − 2)l 1 p
> ||∇ u||2 + ||u||p (21.29)
2p p2
which implies that
2 pd p
> ||∇ u||2 and p 2d > ||u||p . (21.30)
( p − 2)l
It follows from (A1), the definition of J(u), (21.11), and (21.27) that
1 1
∫ u ln|u| dx.
p
p ||ut ||2 + 2||∇ u||pp + pl ||∇ u||2 + ||u||p ≤ 2 pd + p
p p
Ω (21.31)
By using Lemma 21.1, (21.31) shows that
1 p
p ||ut ||2 + 2||∇ u||pp + pl ||∇ u||2 + ||u||p
p
( p − 2)a 2 p a2 p
≤ 2 pd + ||u||p + ||∇ u||p
4π 2π
n p p
− (1 + ln a)||u||p + ||u||p +ln ||u||p . (21.32)
p
Let a = 2π (1 + l ), then by using Lemma 21.6 and (21.30), (21.32) becomes
p ||ut ||2 + (1 − l )||∇ u||pp + pl ||∇ u||2

1
( p − 2)(1 + l )
≤ 2 pd + ||u||pp + ||u||pp ln( p 2d ) p }
2
n 1
− (1 + ln 2π (1 + l ))||u||pp − ||u||pp
p p
p −1 1 (n +1) n
2 pdp  p 
( p − 2)(1+ l )
p p
≤ e + e 2 2
+ ln( p d ) − ln e (2π (1 + l ))  ||u||pp
2p
( p − 1)  
4 pd
= ln p (21.33)
( p − 1)

 1 1 1  1   4 pd
p ||ut ||2 + (1 − l )||∇ u||pp + pl ||∇ u||2 ≤ max  , ,   ln p.
 p 1 − l pl  pl   ( p − 1)
(21.34)
Therefore, we have solutions (u) for the Equation (21.1) that exist globally,
by using (21.34) and by the same argument, the continuation theorem [27, 28].
21.4 Blow Up Results of the Solution

for Equation (21.1)
This section is related to blow up results for Equation (21.1) at infinity. We
state some lemma which will be useful for our proof. Proof Lemma 21.9 can
be obtained by same argument as Lemma 21.7.
Lemma 21.9. Let u be taken as a weak solution of Equation (21.1). Suppose

that u0 ∈ V, u1 ∈ H 01 (Ω ) and 0 < E(0) < d. Therefore, we have u ∈ V and
E(t) < d.
Theorem 21.10. Let u ∈W01,p (Ω ) and u1 ∈ H 01 (Ω ) and ∫ Ω u0u1dx > 0. If
u0 ∈ V and E (0) < d, then the solution u of Equation (21.1) goes to the
infinity when t → ∞, i.e.,
lim u(t ) = ∞.
t →∞
Proof. Under the conditions of Theorem 21.10, we conclude that I(u) is

decreasing with respect to t, thank to results of the Lemma 21.9. Thus, we
can find for all t > 0
t
 
p
I (u) = ||∇ u|| +  1 −
p
 ∫
0
g (s )ds  ||∇ u||2 + ( g °∇ u)(t ) −
 ∫ u ln|u| dx < 0.
Ω
p
(21.35)
By using the Logarithmic Sobolev inequality, (21.35) can be written as

t
 a2   
2π
p
 1 −  ||∇ u||p +  1 −
 ∫ 0
g (s )ds  ||∇ u||2 + ( g °∇ u)(t )

2
( p − 2)a n
− ||u||pp +  (1 + ln a) − ln ||u||p  ||u||pp
4π p 
< 0. (21.36)

2
Since 0 < < 1, by using (21.22), we have
p
 a2 
 1 − + l  (||∇ u||2 −1) + ( g °∇ u)(t )
2π 
n ( p − 2)a 2  p
+  (1 + ln a) − ln ||u||p −  ||u||p
 p 4π 
< 0.
By taking a = 2π (1 + l ) into (21.36) and using definition of d, we intro-

duce that
n ( p − 2)a 2 (21.37)
(1 + ln a) − ln ||u||p − < 0,
p 4π
p 2d < ||u||pp . (21.38)
Now, we set the functional G(t) : [0, ∞) → R+ which was defined by

t
2
G(t ) = ||u|| +
∫ |∇ u(s)| ds + (T − t )||∇ u || .
0
2 0 2
(21.39)
Direct computation on (21.39) leads to
G ′(t ) =
∫ 2uu dx + ||∇ u|| − ||∇ u ||
Ω
t
2
0
2
=
∫ 2uu dx + ∫ ∫ 2|∇ u(s)||∇ u (s)|dxds.
Ω
t
0 Ω
t (21.40)
By using Equation (21.1), we have second-order differentials of G(t) as

follows:
G ′′(t ) = 2
∫ |u | dx + 2 ∫ uu dx + 2 ∫ |∇ u||∇ u |dx
Ω
t
2
Ω
tt
Ω
t
= 2||ut ||2 −2
∫ |∇ u| dx − 2 ∫ |∇ u| dx
Ω
2
Ω
p
+2
∫ ∫ g (t − s)∇ u(s)∇ u(t )dxds + 2 ∫ |u| ln|u|dx.
Ω 0 Ω
p
(21.41)
Since
∫ ∫ g (t − s)∇ u(s)∇ u(t )dsdx

0 Ω
t t
1 
= 
2

∫ 0
2
g (t − s )(||∇ u(t )|| + ||∇ u(s )|| )ds − 2
∫
0
g (t − s )(||∇ u(s ) −∇ u(t )||2 )ds 


t t
1 
= 
2

∫ 0
2
g (t − s )(||∇ u(t )|| )ds − ( g °∇ u)(t ) +
∫ 0
g (t − s )(||∇ u(s )||2 )ds 


(21.41) leads to
t
 
∫
2 p
G ′′(t ) ≥ 2||ut || −2  1 − g (s )ds  ||∇ u||2 dx − 2||∇ u||p
 
0
t
+
∫ g (t − s)(||∇ u(s)|| )ds − 2( g °∇ u)(t ) + 2 ∫ |u| ln |u|dx
0
2
Ω
p
2
≥ 2||ut || −2 I (u). (21.42)
By combination of (21.39), (21.40), and (21.42) we obtain that
( p 2 + 2)
G ′′G − (G ′ )2
4
( t
)
= 2G(t ) ||ut ||2 + ||∇ u||p + 1 − ∫ 0 g (s )ds ||∇ u||2 +( g °∇ u)(t ) + 2 ∫ Ω |u|p ln |u| dx 
p
t
2 2
−( p + 2)(G(t ) − (T − t )||∇ u0 || ) × (||ut || + 2
∫ |∇ u (s)| ds + ( p + 2)(t ).
0
t
2 2
(21.43)
where
t t
   
(t ) =  ||u||2 +
 ∫
0
|∇ u(s )| ds  ×  ||ut ||2 +
 
2
∫ 0
|∇ ut (s )|2ds 

t 2
 
−
 ∫ uu dx + ∫ ∫
Ω
t
0 Ω
|∇ u(s )||∇ ut (s )|dxds .

(21.44)
Thanks to the Cauchy-Schwarz inequality and Young inequality, the last

term of (t ) can be rewritten as
∫ uu dx ≤ ||u|| ||u || ,
Ω
t
2
t
2
(21.45)
t t t
∫ ∫ |∇ u(s)||∇ u (s)|dxds ≤ ∫ ||∇ u(s)|| ds ∫ ||∇ u (s)|| ds,

0 Ω
t
0
2
0
t
2
(21.46)
and
t
2
∫ uu dx + ∫ ∫ |∇ u(s)||∇ u (s)|dxds
Ω
t
0 Ω
t
t t
≤ ||ut || 2
∫ ||∇ u(s)|| ds + ||u|| ∫ ||∇ u (s)|| ds.
0
2 2
0
t
2
(21.47)
Moreover, by inserting (21.45)-(21.47) into (21.44), it is clearly seen that

(t ) ≥ 0. Depending on this interest, we replace (21.43) as
p2 + 2
G ′′G − (G ′ )2 ≥ ΛG (21.48)
4
where Λ = Λ(t) is defined such that
  t
 
Λ(t ) = 2 ||ut ||2 + ||∇ u||p
∫ ∫ |u|p ln|u|dx 
p
+1 − g (s )ds  ||∇ u||2 + ( g °∇ u)(t ) −
   
 0 Ω 
t
 
− ( p + 2) ||ut ||2 +
 ∫
0
|∇ ut (s )|2 ds  .

By using definition of total energy E(t), (A1), (A2), (21.22) and the
Logarithmic Sobolev inequality, it follows that
Λ(t ) ≥ −2 pE(t ) + 2( p + 1)||∇ u||pp + 2||u||pp }

t
 
2
+( p + 2) 1 −
 ∫0
g (s )ds  ||∇ u||2 + ( p 2 + 2)( g °∇ u)(t )

t
−2( p + 1)
∫ |u| ln|u|dx − ( p + 2)∫ |∇ u (s)| ds
Ω
p 2
0
t
2
2
 a  p
≥ −2 p 2 E(t ) + 2( p + 1) 1 −  ||∇ u||p
 2π 
t
 
2
+( p + 2) 1 −
 ∫0
g (s )ds  ||∇ u||2 +( p 2 + 2)( g °∇ u)(t )

 ( p − 2)a 2 n  p
−2( p + 1)  − (1 + ln a) + ln ||u||p  ||u||p
 4π p 
t
∫ |∇ u (s)| ds + 2||u||
2 2 p
−( p + 2) t p (21.49)
0
 a2  p
≥ −2 p 2 E(t ) + ( p 2 + 2) 1 −  + l )(||∇ u||2 −1) + 2||u||p
 2π 
 ( p − 2)a 2 n  p
−2( p + 1)  − (1 + ln a) + ln ||u||p  ||u||p
 0 
 ( p − 2)a 2 n  p
−2( p + 1)  − (1 + ln a) + ln ||u||p  ||u||p
 4π p 
t
∫
2 2 p
322 Mathematics |∇ utComputer
−( p + 2) and (s )| ds + 2||u||p Volume 2
Science
0
 a2  p
≥ −2 p 2 E(t ) + ( p 2 + 2) 1 −  + l )(||∇ u||2 −1) + 2||u||p
 2π 
 ( p − 2)a 2 n  p
−2( p + 1)  − (1 + ln a) + ln ||u||p  ||u||p
 4π p 
t
2
−( p + 2)
∫ |∇ u (s)| ds.
0
t
2
By p > 2, a = (2π (1 + l ), (21.37), (21.38), and (21.15), (21.49) can be

found as
t
2
Λ(t ) ≥ −2 p E(t ) − ( p + 2)2
∫ |∇ u (s)| ds + 2 p d
0
t
2 2
t t
 
2
≥ −2 p  E(0) −
 ∫ 0
|∇ ut (s )| ds  + 2 p 2d − ( p 2 + 2)
2
 ∫ |∇ u (s)| ds
0
t
2
= −2 p 2 E(0) + 2 p 2d + ( p 2 − 2)
∫ |∇ u (s)| ds.
0
t
2
(21.50)
Therefore, we obtain from the condition of Theorem 21.10 (E (0) < d)
Λ(t) ≥ 0. (21.51)
Therefore, by Inequality (21.51) and condition of the functional G(t),

(21.48) yields that
p2 + 2
G′′G − (G′)2 ≥ 0. (21.52)
4
By directly calculation of (21.52), it follows that
p2 + 2
ln|G ′(t )| − ln|G(t )| 4 ≥ 0. (21.53)
By integrating both sides of Inequality (21.53) over [0, t] for p > 2, it

yields that
G(0)
G(t ) ≥ 4 . (21.54)
 p2 − 2 p2 − 2  2
p −2
1 − G(0) 4 t
 4 
p2 − 2
 G ′(0)  4
From definition of G(t), it is clear to conclude that  < 1 where
 G(0) 
G(0) =|| u ||2 + T || ∇u ||2 . Consequently, we can rewrite (21.54) so that
G(0)
G(t ) ≥ 4 .
2
 ( p − 2)  G ′(0)   p2 − 2
1 − 4  G(0)  

Therefore, lim G(t ) = ∞ for some T* satisfies

t →T *
4  G(0)  4||u||2 + 4T ||∇ u0 ||2

0 <T* < ≤
p 2 − 2  G ′(0)  2( p 2 − 2) ∫ Ω uu0t dx
which means that
2||u||2
T≤
( p 2 − 2) ∫ uu dx − 2||∇ u || .
Ω
0t 0
2
so that our solution blows up at T*. This finishes the proof.

References
1. Adams, R. A., Fournier, J. J. Sobolev spaces. Pure and applied mathematics
(Amsterdam), (2003).
2. Al-Gharabli, M. M., Al-Mahdi, A. M., Messaoudi, S. A. Decay Results for a
Viscoelastic Prob- lem with Nonlinear Boundary Feedback and Logarithmic
Source Term. Journal of Dynamical and Control Systems, 1-19, (2020).
3. Andrade, D., Silva, M. J., Ma, T. F. Exponential stability for a plate equation
with p-Laplacian and memory terms. Mathematical Methods in the Applied
Sciences, 4(35), 417-426, (2012).
4. Araújo, G. M., Araújo, M. A. F., Pereira, D. C.On a variational inequality for
a plate equation with p-Laplacian end memory terms. Applicable Analysis,
1-14, (2020).
5. Bialynicki-Birula, I., Mycielski, J. Wave equations with logarithmic nonlin-
earities. Bull. Acad. Pol. Sci. Cl, 3(23), 461-466, (1975).
6. Bouhali, K., Ellaggoune, F. Viscoelastic wave equation with logarithmic non-
linearities in Rn. J Partial Differ Equ, 30(1), 47-63, (2017).
7. Buljan, H., Šiber, A., Soljačíc, M., Schwartz, T., Segev, M., Christodoulides,
D. N. Incoherent white light solitons in logarithmically saturable noninstan-
taneous nonlinear media. Physical Review E, 68(3), 036607, (2003).
8. Cazenave, T., Haraux, A. Équations d’évolution avec non linéarité logarith-
mique. In Annales de la Faculté des sciences de Toulouse: Mathématiques,
2(1), 21-51, (1980).
9. Di, H., Shang, Y., Song, Z. Initial boundary value problem for a class of
strongly damped semilinear wave equations with logarithmic nonlinearity.
Nonlinear Analysis: Real World Ap- plications, 51, 102968, (2020).
10. Ding, H., Wang, R., Zhou, J. Infinite time blow-up of solutions to a class of
wave equations with weak and strong damping terms and logarithmic non-
linearity. Studies in Applied Mathematics, 147(3), 914-934, (2021).
11. Ferreira, J., Irkıl, N., Raposo, C. A. Blow up results for a viscoelastic
Kirchhoff-type equation with logarithmic nonlinearity and strong damping.
Mathematica Moravica, 25(2), 125-141, (2021).
12. Ha, T. G., Park, S. H. Blow-up phenomena for a viscoelastic wave equation
with strong damping and logarithmic nonlinearity. Advances in Difference
Equations, 2020 (1), 1-17, (2020).
13. Liu, W., Yu, J. On decay and blow-up of the solution for a viscoelastic wave
equation with boundary damping and source terms. Nonlinear Analysis:
Theory, Methods Applications, 74(6), 2175-2190, (2011).
14. Liu, W., Li, G., Hong, L. Decay of solutions for a plate equation with
p-Laplacian and memory term. Electronic Journal of Differential Equations,
2012(129), 1-5, (2012).
15. Liu, C., Ma, Y. Blow up for a fourth order hyperbolic equation with the loga-
rithmic nonlinearity. Applied Mathematics Letters, 98, 1-6, (2019).
16. Ma, L., Fang, Z. B. Energy decay estimates and infinite blow-up phenomena
for a strongly damped semilinear wave equation with logarithmic nonlinear
source. Mathematical Methods in the Applied Sciences, 41(7), 2639-2653,
(2018).
17. De Martino, S., Falanga, M., Godano, C., Lauro, G. Logarithmic Schrödinger-
like equation as a model for magma transport. EPL (Europhysics Letters),
63(3), 472, (2003).
18. Park, S. H. Stability for a viscoelastic plate equation with p-Laplacian. Bulletin
of the Korean Mathematical Society, 52 (3), 907-914, (2015).
19. Pereira, D. C., de Araújo, G. M., Raposo, C. A.. Unilateral Problem for a
Viscoelastic Beam Equation Type p-Laplacian with Strong Damping and
Logarithmic Source. European Journal of Mathematical Analysis, 2 (5), 1-20,
(2022).
20. Peyravi, A. General stability and exponential growth for a class of semi-
linear wave equations with logarithmic source and memory terms. Applied
Mathematics and Optimization, 81(2), 545-561, (2020).
21. Del Pino, M., Dolbeault, J., Gentil, I. Nonlinear diffusions, hypercontractiv-
ity and the optimal Lp-Euclidean logarithmic Sobolev inequality. Journal of
Mathematical Analysis and Applications, 293(2), 375-388, (2004).
22. Pişkin, E., Okutmuştur, B. An Introduction to Sobolev Spaces, Bentham
Science, (2021).
23. Pişkin, E., Irkıl, N. Well-posedness results for a sixth-order logarithmic
Boussinesq equation. Filomat, 33(13), 3985-4000, (2019).
24. Pişkin, E., Boulaaras, S., Irkil, N. Qualitative analysis of solutions for the p-
Laplacian hyperbolic equation with logarithmic nonlinearity. Mathematical
Methods in the Applied Sciences, 44(6), 4654-4672, (2021).
25. Irkıl, N., Pişkin, E. Global Existence and Decay of Solutions for a Higher-
Order Kirchhoff- Type Systems with Logarithmic Nonlinearities. Quaestiones
Mathematicae, 1-24, (2021), (in press).
26. Raposo, C. A., Cattai, A. P., Riberio, J. O.Global solution and asymptotic
behaviour for a wave equation type p-Laplacian with memory. Open Journal
of Mathematical Analysis, 2, 156-171, (2018).
27. Segal, I. Nonlinear semigroups. Annals of Mathematics, 339-364, (1963).
28. Georgiev, V., Todorova, G. Existence of a solution of the wave equation with
nonlinear damping and source terms. Journal of differential equations, 109
(2), 295-308, (1994).
29. M. Willem, Minimax Theorems. Progress in Nonlinear Differential Equations
and Their Applications 24, Birkhouser Boston Inc. Boston, MA, (1996).
30. Xu, R., Lian, W., Kong, X., Yang, Y. Fourth order wave equation with non-
linear strain and logarithmic nonlinearity. Applied Numerical Mathematics,
141, 185-205, (2019).
31. Ye, Y. Logarithmic viscoelastic wave equation in three-dimensional space.
Applicable Analysis, 100(10), 2210-2226, (2021).
32. Ye, Y.Global solution and blow-up of logarithmic Klein-Gordon equation.

Bulletin of the Korean Mathematical Society, 57(2), 281-294, (2020).
33. Yaojun, Y. E. Existence and nonexistence of global solutions for logarithmic
hyperbolic equation. Authorea Preprints, (2021).
34. Zhou, J. Behavior of solutions to a fourth-order nonlinear parabolic equa-
tion with logarithmic nonlinearity. Applied Mathematics and Optimization,
84(1), 191-225, (2021).
35. Zu, G., Guo, B. Bounds for lifespan of solutions to strongly damped semi-
linear wave equations with logarithmic sources and arbitrary initial energy.
Evolution Equations and Control Theory, 10(2), 259, (2021).
22
Flow Dynamics in Continuous-
Time with Average Arc Capacities
Badri Prasad Pangeni1* and Tanka Nath Dhamala2
Tribhuvan University, Prithvi Narayan Campus, Pokhara, Nepal

1
2
Tribhuvan University, Central Department of Mathematics, Kathmandu, Nepal
Abstract
The dynamics of network flows are influenced by transit time along the arcs, unlike
the static cases. A continuous-time model depicts the actual scenario of many
transshipment problems. Here, we introduce the average arc problem (AAP) in a
continuous-time model which widely occurs in our daily life. In AAP, the flow value
which is total of the flow at every time step/instant (in discrete/continuous time
respectively) of the arc has an upper bound as the flow capacity of that arc. When
the time-expanded network is structured, AAP violates the capacity constraint. This
failure of extension of the arc problem (AP) to the AAP of the problem leads to for-
mulating a linear programming of it. In AP, at every instant of time, flow can be sent
up to the capacity of that arc, as in the traditional dynamic network flows. To get
feasible flow, a transformation and reduction algorithm is also presented.
Keywords: Dynamic network flows, continuous-time model, complexity,

linear programming
22.1 Introduction
Consider the water supply network of a city, the traffic network of a high-
way connecting two cities, or the communication network between two
places. In these type of network flows, the capacity of the arc in every time
step in discrete time (in every instant in continuous-time) can be fulfilled
by the corresponding flows. Ford and Fulkerson [6] introduced such flow
dynamics by making dynamic networks with a source and destination.
*Corresponding author: badripangeni@gmail.com
327
The algorithm with polynomial running time was firstly studied by Hoppe
and Tardos [10] and solved flow sending problems in dynamic cases hav-
ing more than one source and/or sink.
In such cases, if the transit time and capacity of an arc are 5 and 3 units
(say) respectively, then each time step can contain 3 units of flow, as shown
in Figure 22.1. Here, the capacity is not fulfilled in time steps 1 and 4. We
refer to this type of flow problem as the arc problem (AP).
In our paper, we have switched to the cases where some restrictions in
each time steps flow cannot be fulfilled. Total value of the flow of all time
steps of the transit time of an arc is bounded above by the capacity of that
arc (see Figure 22.2), where total of the flow value of time steps 1, 3, and 4
is 3, the capacity of that arc. We call this problem the average arc problem
(AAP).
The structure of this paper is as follows. Section 22.2 gives background
on how flow dynamics work in a discrete as well as continuous-time set-
ting along with the short view on AAP. In Section 22.3, failure of extension
of AP to AAP is discussed with the help of a time-expanded network. A
mathematical formulation for the AAP and transformation and reduction
algorithm are introduced in Section 22.4. Lastly, the paper is wrapped with
conclusion in Section 22.5.
0 1 2 3 4 5
Figure 22.1 Arc with transit time 5 and capacity 3 units.
0 1 2 3 4 5
Figure 22.2 Average arc with transit time 5 and capacity 3 units.
Flow Dynamics in Continuous-Time with Average Arc Capacities 329
22.2 Literature Review

Ford and Fulkerson introduced the notion of network flows in dynamic
cases [5, 6]. Gale [7] presented the earliest arrival flow problem. An example
in the case of only one source and two sinks having no earliest arrival flow
was given by Baumann and Skutella [3]. The quickest flow problem with only
one source and single sink related to the maximum flow problem in dynamic
status was used by Burkard et al. [4] and also proposed a polynomial as well
as strongly polynomial algorithms for the quickest network flow problem.
Dynamic network flow problems with polynomial time complexity were
presented by Hoppe and Tardos [9]. Aronson [2] surveyed the network flow
problems on discrete dynamic cases, especially on the flow having maxi-
mum value. Hamacher and Tjandra [8] focus mainly on building evacu-
ation. Kotnyek [12] also gives a review on dynamic network flows. Kohler
et al. [11] introduces the congestion concept on static and dynamic flows.
Network flow problems in continuous-time setting to get maximum
flow value were first introduced by Philpott [17] and further extended by
Anderson, Nash, and Philpott [1], where they consider the problem with-
out transit times and variable transit and storage capacities. This result
with transit time on arc was later extended by Philpott [16].
Skutella [18] presents a survey of dynamic network flows in continu-
ous-time setting with constant transit times and capacities. Nasrabadi and
Hashemi [14] investigate a dynamic flow problem having minimum cost
with an algorithm where time values are discrete. Pangeni and Dhamala
[15] reviewed on dynamic network flows in continuous-time model.
A joint project of the Dahlgren Lab of the U.S. Navy with Cornell
University [13] first raised the AAP type problem with the aim to execute
an algorithm for solving the AP [10] and to test it on an available data set.
22.3 Failure in Extension of AP to AAP

Consider a directed network N = (V, A) having a set of vertices or nodes
V and set of arcs A, such that A ⊂ V × V, i.e., vertices are the crossing-over
of arcs.
Let fij : A× [0, T ] → R≥0 be a continuous-time dynamic flow in arc (i, j)
over the time horizon [0, T ], where fij(t) is the rate of flow entering the arc
(i, j) in time t. Also, let uij(t), wij(t), and τij(t), respectively, be the capacity
of the flow rate, cost of unit flow, and traversal time in arc (i, j) at time t.
Let si : V × [0, T ] → R≥0 be a continuous-time dynamic storage at ver-
tex i over [0, T ]. Also, let vi(t), xi(t), and mi(t), respectively be the storage
capacity, per time unit storage cost per unit flow, and supply/demand rate
at vertex or node i, at time t.
Let Fi(t) be the flow rate from supersource ssi to source si(from sink di to
supersink sdi) at time t. Also, let Fi be the flow amount supplied from super-
supersource sss to supersource ssi (from supersink sdi to supersupersink ssd).
Example. Consider an instance of an AP given in Figure 22.3 and its time-

expanded network, respecting the transit times of each arc in Figure 22.4.
sss
ss1 ss2
m=6 s1 s2 m = 4
2
1 1
1
x y
2 1
d m = 10
sd
ssd
Figure 22.3 Simple instance of dynamic network of an arc problem, arc with transit time,
and capacity 1 unit each.
s1 s2 x y d 3
ss2 s1 s2 x y d [2, 3)
sss sd ssd
ss1 s1 s2 x y d [1, 2)
s1 s2 x y d [0, 1)
Figure 22.4 Time-expanded network of Figure 22.3 over T = 3.

Consider the flow through the arc (s1, y) in time [0, 1) and [1, 2] and
the total flow value is 1 + 1 = 2 units, which violates the capacity of flow in
AAP which leads to infeasibility of flow and hence the failure in extension
of AP to AAP.
22.4 Formulation
Using the notations introduced above, the minimum time of dynamic flow
problem (MTP) suggested by Philpott [16] is:
MTP:
min T (22.1)
∫ ∑ ∫ ∑ f (t −τ ) + s (t ) =
t t
s.t . fij (t )dt − ji s (0) ji i i (22.2)
0 0
0 ≤ fij ≤ uij, 0 ≤ si ≤ vi, s(t) = 0, t ≥ T, fij(t) = 0, t ≥ T − τij (22.3)
The linear programming formulation for AAP in a discrete-time model

formulated by Melkonian [13] is reformulated in a continuous-time model
by the authors as follows:
AAP:
∫ ∑ ∑ (t + τ
T
min id ) fid (t ) dt (22.4)
0
i →d t∈[0,T ]
t +τ ij
∫ ∑ f (t ) dt ≤ u
t
ij ij (22.5)
0
t
fij(t) ≤ uij (22.6)
Fi(t) ≤ mi (22.7)
Fi ≤ mi (22.8)
∫ ∑ ∫ ∑f
t t
fij (t − τ ij ) dt = jk (t ) dt (22.9)
0 0
i→ j j→k
∫ ∑ ∫ ∑f
t t
− τ ij ) dt + s j (0)
fij (t= jk (t ) dt + s j (t ) (22.10)
0 0
i→ j j→k
∫ ∑ f (t ) dt
t
Fs (t ) = sj (22.11)
0
s→ j
∫ ∑ f (t −τ
t
=Fd (t ) id id ) dt (22.12)
0
i →d
∫ ∑ F (t ) dt
T
Fi = i (22.13)
0
i∈ssi /sdi
∑F
s∈sources
s = ∑m
s∈sources
s (22.14)
∑F
d∈sinks
d = ∑m
d∈sinks
d (22.15)
The linear objective function 22.4 aims to minimize the arrival time
of the flows at the sink d. Similarly, 22.5, 22.6, 22.7, and 22.8 give capac-
ity constraints of the flows, 22.9 is the flow conservation constraint for
intermediate vertex j, 22.10 expresses flow non-conservation constraint,
22.11 and 22.12, respectively, are the conservation constraints for flow in
the sources and sinks, 22.13 is the flow conservation constraints for the
supersources and supersinks, and lastly 22.14 and 22.15 are the total mass
supplied/demanded from supersupersource/supersupersink.
The feasibility of AAP and the above given linear formulation are
bi-conditional to each other. To observe the feasibility of flow, the concept
of PARTITION is used.
PARTITION Problem
Consider a finite set which has n items with sizes b1, ..., bn ∈ Z+, such that
∑nk =1 bk =
2 J , for some J ∈ Z+.
Does some subset I ⊂ {1, ..., n} satisfy the property ∑ k∈I bk = J?
Theorem 22.1 [13] The necessary and sufficient condition to send two units
of dynamic flow from s to d in time T is the existence of a ‘yes’-instance
PARTITION problem.
The idea to prove the sufficient part of Theorem 22.1 is the flow of two
units is sent along two paths having a disjointed arc in two packets, each
having one unit of flow and length n and this can be done in both cases
within the prescribed time horizon T. For the necessary part, firstly
an arbitrary arc is taken which has more than one unit flow value and
total time required to reach the flow to the sink d is calculated which
yields more than the time horizon T, which leads to the contradiction.
Again, if a supposition is made against the yes instance PARTITION, i.e.,
∑ k∈I bk = J + ∈, for some ϵ ≠ 0, then total transit time of the flow on one
of the two paths will exceed T.
If an NP-complete problem fails to be NP-complete after satisfying sim-
ilarity assumption, then it is called weakly NP-complete. Using the concept
of PARTITION and the above theorem, Melkonian [13] showed AAP is
NP-complete in the weak sense. To reduce AAP to AP, only the capacities
of the arcs are modified, dividing by the corresponding transit times of
the arcs where capacity constraints are violated. The following algorithm
performs well for the feasibility of flow with the time optimization given
by the AAP formulation and its execution steps ahead on the basis of the
length of arc having the least transit time.
ALGORITHM 22.1 Transformation and Reduction
- Input: Given an AP with N = (V, A, T, fij, uij, wij, τij, si, vi, xi, mi)
1. Transform AP to time-expanded network

2. If fij ≤ uij in AAP, do. Otherwise, go to Step 3
3. If fij > uij in AAP, go to Step 4 (capacity violation)
4. Formulate AAP
5. Use NP-complete PARTITION problem to reduction
6. Reduce AAP to AP by averaging the capacity of arcs where
violation occurs
7. Feasibility of AP implies the feasibility of AAP [10]
8. Construct hybrid network, i.e, mixed capacity network
- Output: Feasible flow

22.5 Conclusion
When AP is transformed into the time-expanded network, it violates
capacity constraints in AAP. This infeasibility of flow leads to the linear
formulation of AAP. Using the well-known NP-complete PARTITION
problem, AAP is reduced to AP by some modification on arc capacities, i.e,
averaging the capacities by the corresponding transit times. The feasibility
of reduced AP can be obtained by the algorithm introduced by Hoppe and
Tardos [10], implying the feasibility of the original AAP. Practically, the
hybrid network would be functioning simultaneously. After observing the
feasibility of the flow, the optimality can be obtained by the suitable opti-
mality seeking method (such as network simplex method, etc.).
Acknowledgment
The first author acknowledges University Grants Commission, Nepal for
PhD fellowship.
References
1. E. J. Anderson, P. Nash and A. B. Philpott (1982). A class of continuous net-
work flow problems. Mathematics of Operation Research, 7: 501-514.
2. J. E. Aronson (1989). A survey of dynamic network flows. Annals of
Operations Research, 20: 1-66.
3. N. Baumann and M. Skutella (2006). Solving evacuation problems effi-
ciently–earliest arrival flows with multiple sources. In: 47th annual IEEE
symposium on foundations of computer science (FOCS’06), 399–410.
4. R. Burkard, K. Dlaska and B. Klinz (1993). The quickest flow problem. ZOR
Meth Models Oper Res, 37: 31–58.
5. L. R. Ford and D. R. Fulkerson (1958). Constructing maximal dynamic flows
from static flows. Operations Research Letters, 6: 419-433.
6. L. R. Ford and D. R. Fulkerson (1962). Flows in networks. Princeton
University Press, New Jersey.
7. D. Gale(1959). Transient flows in networks. Michigan Math J, 6(1): 59–63.
8. H. Hamacher and S. Tjandra (2001). Mathematical modeling of evacuation
problems: a state of art. Berichte des Frauenhofer ITWM, Nr. 24. http://
www.itwm.fraunhofer.de/fileadmin/ITWMMedia/Zentral/Pdf/Berichte
ITWM/2001/bericht24.pdf. Accessed 18 Oct 2006.
9. B. Hoppe and E. Tardos (1994). Polynomial time algorithms for some evacu-
ation problems. In: Proceedings of the fifth annual ACM-SIAM symposium
on discrete algorithms, Society for Industrial and Applied Mathematics

Philadelphia, PA, 433–441.
10. B. Hoppe and É. Tardos (2000). The quickest transshipment problem.
Mathematics of Operations Research 25: 36–62.
11. E. Kohler, R. H. Mohring, and M. Skutella (2009). Traffic networks and flows
over time. In: Lerner J, Wagner D, Zweig K (eds) Algorithmics of large and
complex networks. Lecture notes in computer science, vol 5515. Springer,
Berlin, 166–196.
12. B. Kotnyek (2003). An annotated overview of dynamic network flows. INRIA,
Sophia Antipolis, France.
13. V. Melkonian (2007). Flows in dynamic networks with aggregate arc capaci-
ties. Information Processing Letters 101: 30–35.
14. E. Nasrabadi and S. M. Hashemi (2010). Minimum cost time-varying net-
work flow problems. Optim Meth Software, 25(3): 429–447.
15. B. P. Pangeni and T. N. Dhamala (2021). A brief survey on dynamic net-
work flows in continuous-time model. Journal of Mathematical Sciences and
Computational Mathematics 2(4): 467-477.
16. A. B. Philpott (1990). Continuous-time flows in networks. Mathematics of
Operation Research, 15: 640-661.
17. A. B. Philpott (1982). Algorithms for continuous network flow problems.
Ph.D. thesis, University of Cambridge UK.
18. M. Skutella (2009). An introduction to network flows over time. In W.
Cook, L. Lovasz and J. Vygen (eds.): Research Trends in Combinatorial
Optimization, Springer, Berlin, 451-482.
23
Analysis of a Multiserver System
of Queue-Dependent Channel
Using Genetic Algorithm
Anupama1* and Chandan Kumar2
Department of Mathematics, Darbhanga College of Engineering, Darbhanga,

1
Bihar, India
2
Department of Mechanical Engineering, BIT Sindri, Jharkhand, India
Abstract
In the present work, a M/M/3 Queuing system with a Queue-dependent multi-server
has been considered. Here, a number of failed machines form a queue and repair-
men consider a service provider or service channel which starts its service when
the queue length is N. We found a generating function for breakdown machines.
Afterwards, some performance measures including idle time and busy time for the
system have been evaluated. At last, cost is optimized using a genetic algorithm.
Keywords: M/M/3, repairmen, queue dependent server, genetic algorithm
23.1 Introduction
The concept of a multiserver queueing system with queue-dependent serv-
ers is not new. Hahn and Sivazlian, in 1990, developed the M/M/2 model
with service stations [5]. Yadin and Naor, in 1963, invented the N policy
concept for the single server system [9]. The (0, K, N, M) rule in a two
server system used by Rhee and Sivazlian in 1990 helps to derive the work-
ing period distribution [12].
Natarajan studied a system for warm stand-bys in 1968 [13]. With one
online unit and two standbys with failure and repair time exponentially
distributed, this model assumes only one condition, namely that the sys-
tem will fail if spares are not present for the breakdown machine. There is
*Corresponding author: singhlovely82@gmail.com
337
an analytical closed-form solution for the M/Ek/1 system [15]. The server
system for standbys and 2 types of failure time distribution is also examined.
Wang and Ke (2000) introduce a supplementary and recursive technique
for developing a solutions to a single server queueing system working with
the N policy [16]. The current system is that the line length determines the
server value in the system. In the event that no machines fail in the system,
all the fixers are temporarily inactive until the length of the line reaches a
predetermined level, i.e., W, which is also a decision change. For this sin-
gle server period, it will be activated instantly. Sometime later, if the line
length in the system goes up to the aforementioned level, it says S, which is
less than W and then the other available server will also run immediately.
Similarly, if the failure machines in the system go up to a next level, say T,
which is less than S, then all the servers will be activated immediately.
The objectives of this paper are:
i. To achieve a sustainable distribution of waiting line

opportunities
ii. Finding average count of failed machines
iii. To design the total cost of the model
iv. The genetic algorithm method is used to optimize costs
23.2 Description of the Model

The states for the model depicted by [i, n], i = 0, 1, 2, 3 represent that no
repairman is working, one of the repairmen is working, two of the repair-
men working, or three of the repairmen working, respectively. n ϵ [0, L] is
the number of failed machines in the model.
23.3 Notations
ϸ[0, n] = Occurrence of n customers in the model where there are
no operators, when n ϵ [0, W-1]
ϸ[1, n] = Occurrence of n customers in the model when one mod-
ifier works, when n ϵ [0, S-1]
ϸ[2, n] = Occurrence of n customers in the model, where two
moderators are working, when n ϵ [q + 1, T-1]
Analysis of a Multiserver System of Queue-Dependent Channel 339
ϸ[3, n] = Probability of n customers in the system, where three

moderators are working, when n ϵ [ r + 1, L].
α = Average arrival rate
β = Average service level
ϸ0 = Zero server probability in the system
ϸ1 = One server probability in the system
ϸ2 = Two server probability in the system
ϸ3 = Three server probability in the system
L0 = Line length of the system if all fixers are on vacation
L1 = Length of the line in the system where a single server is
working in the system
L2 = Line length in the system where two modifiers are working
in the system
L3 = Line length in the system where all three fixers work in the
system
Ls = Line length in system
EI =Idle time expected value
E B1 = Busy time expected length during which any repairman is
working on the system
E B2 = Busy time expected length during which any repairman is
working on the system
E B3 = Expected length of busy time during which every repair-
man works on the system
E C = Busy time expected length
Ch = Cost of each failed machine during the unit available in
the system
Cv1 = Cost savings per failed machine during each unit in the
system when 1 mechanic is on vacation
Cv2 = Cost of handling per failed machine during each unit
present in the system when 2 technicians are on vacation
Cv3 = Cost savings per failed machine during each unit present
in the system when 3 technicians are on vacation
C01 = Holding cost of single technician functioning in the sys-
tem during each unit time
C02 = Holding cost of 2 technicians functioning in the system
during unit time
C03 = Holding cost of 3 technicians functioning in the system
during each unit time
23.4 Steady State Equations

A diagram of the rate of change is discussed in Anupama and Solanki (Nov
2014). A diagram of the arrival and departure of the failed machines is
shown. Also, server installation is based entirely on line length in the spec-
ified system.
The steady state probability equations for the system are given below:
α ∗ ϸ[0,0] = β ∗ ϸ[1,1] (23.1)
α * ϸ[0, n] = α * ϸ[0, n–1] n є [1, W – 1] (23.2)
[α + β] ∗ ϸ[1,1] = β ∗ ϸ[1,2] (23.3)
[α + β] * ϸ[1, n] = α * ϸ[1, n–1] + β * ϸ[1, n+1] n є [2, q – 1] (23.4)
[α + β] ∗ ϸ[1,q] = α ∗ ϸ[1,q−1] + β ∗ ϸ[1,q+1] + 2 ∗ β ∗ ϸ[2,q+1] (23.5)
[α + β] * ϸ[1, n] = α * ϸ[1, n–1] + β * ϸ[1, n+1] n є [q + 1, W – 1] (23.6)
[α + β] ∗ ϸ[1,W] = α ∗ ϸ[1,W−1] + β ∗ ϸ[1,W+1] + α ∗ ϸ[0,W−1] (23.7)
[α + β] * ϸ[1, n] = α * ϸ[1, n–1] + β * ϸ[1, n+1] n є [W + 1, S – 2] (23.8)
[α + β] ∗ ϸ[1,S−1] = α ∗ ϸ[1,S−2] (23.9)
[α + 2 β] ∗ ϸ[2,q+1] = 2 ∗ β ∗ ϸ[2,q+2] (23.10)
[α + 2β] * ϸ[2, n] = α * ϸ[2, n–1] + 2 * β * ϸ[2, n+1] n є [q + 2, r – 1] (23.11)
[α + 2 β] ∗ ϸ[2,r] = α ∗ ϸ[2,r−1] + 2 ∗ β ∗ ϸ[2,r+1] + 3 ∗ β ∗ ϸ[3,r+1] (23.12)
[α + 2β] * ϸ[2, n] = α * ϸ[2, n–1] + 2 * β * ϸ[2, n+1] n є [r + 1, S – 1] (23.13)
[α + 2 β] ∗ ϸ[2,S] = α ∗ ϸ[2,S−1] + 2 ∗ β ∗ ϸ[2,S+1] + α ∗ ϸ[1,S−1] (23.14)

[α + 2β] * ϸ[2, n] = α * ϸ[2, n–1] + 2 * β * ϸ[2, n+1] n є [S + 1, T – 2] (23.15)
[α + 2 β] ∗ ϸ[2,T−1] = α ∗ ϸ[2,T−2] (23.16)
[α + 3 β] ∗ ϸ[3,r+1] = 3 ∗ β ∗ ϸ[3,r+2] (23.17)
[α + 3β] * ϸ[3, n] = α * ϸ[3, n–1] + 3 * β * ϸ[3, n+1] n є [r + 2, T – 1]

(23.18)
[α + 3 β] ∗ þ[3,T] = α ∗ þ[3,T−1] + 3 ∗ β ∗ þ[3,T+1] + α ∗ þ[2,T−1] (23.19)
[α + 3β] * þ[3, n] = α * þ[3, n–1] + 3 * β * þ[2, n+1] n є [T + 1, L – 1]

(23.20)
α ∗ þ[3,L−1] = 3 ∗ β ∗ þ[3,L] (23.21)
The probability ϸ[i,n] (i = 0-3) from equation 23.1-23.21 are found by

using recursive method
þ[0, n] = þ[0,0], n є [1, W – 1] (23.22)
[1 − n ]
 [1,n] =  [0,0] , n  [1, q] (23.23)
(1 − )
 (1 − n ) (S − S−W )(1 − n−Q ) 

 [1,n] =  +   [0,0] n  [q + 1, W]
 (1 − ) (1 − )(1 − S−Q ) 
(23.24)
 (1 − n )(n−W − S−W ) 

 [1,n] =  S−Q  [0,0] , n  [W + 1, S −1] (23.25)
 (1 − )(1 −  ) 
S−W −1 W
   n−Q 
 (1 −  ) 1 −   
 [ 2 ,n ] =   2   , n  [q +1, S] (23.26)
[0,0]
(2 − )(1 − S−Q )
   
n−S S−Q
S−W −1 W
 (1 −  )   1 −   
2  2 
 [ 2 ,n ] = [0 ,0] , n  [S, S + 1]
(2 − )(1 − S−Q )
(23.27)
T−S−1 S
   n−r 
 (1 −  ) 1 −   
 [3 ,n] =   3    , n  [r +1,T ] (23.28)
[0 ,0]
(3 − )(1 − ρ T−r )
   
n −T T−r
T−S−1 S
 (1 −  )   1 −     [0 ,0]
3  3 
 [ 3 ,n ] = , n  [ T + 1, L] (23.29)
(3 − )(1 − T−r )
α
Where  =
β
23.4.1 Performance Characteristics of the System
0 = zero server probability = ∑ nW=−01 [0,n] ,

S −1
1 = One server probability = ∑

n =1
[1,n] ,
T −1
2 = Two server probability = ∑

n =Q +1
[2,n] ,
3 = Three server probability = ∑

n = r +1
[3,n] ,
23.4.2 Queue Length Evaluations at Different Epochs
W −1
L 0 = Length of queue when all server are on vacation = ∑ n

n =1
[0,n] ,
S −1
L1 = Length of queue when one server is working in the system = ∑ n

n =1
[1,n] ,
T −1
L 2 = Length of queue when two server is working in the system = ∑ n

n = Q +1
[2,n] ,
L 3 = Length of queue when one server is working in the system = ∑ n

n = r +1
[3,n]
23.4.3 Leisure Period and Working Period Length
Expected Leisure period , EI = W / α (23.30)
The queue length of a busy one-time server one, busy two-time server, and
busy three-time server are calculated. Since a working period is the addi-
tion of the leisure time, the busy time of 1 fixer, the busy time of 2 fixers,
and the busy time of 3 fixers, EC = EB1 + EB2 + EB3 + EI
23.4.4 Cost of the System

Total expected cost during each unit of the M \ M \ 3 line system with a lim-
ited power L is measured. Here, {0, (q, r), W, S, T} are variable resolutions. In
line with our goal of determining the maximum number of control variables,
{(q, r), W, S, T} states (q*, r*, W*, S*, T*) to decrease the cost of the model.
The cost of the model is below.
Expected cost ,E{(q , r ),W , S,T }

EI EB EB
= Ch Ls + (CV 1 + CV 2 + CV 3 ) + C01 1 + C02 2
EC EC EC
EB3
+ C03 + α Cl P (3, l) (23.31)
EC

Since the total amount of expected cost of work is unstructured, it is

very tough to the correct values by analyzing reduction in the total cost of
the expected cost over time for each unit. A specific search method can be
used to find important results. In a search method, it is mandatory to have
high limits for dynamic decisions and the search method can be discontin-
ued after a complete global search in the internal area. The upper bound is
L for the optimal values (q*, r*, W*, S*, T*).
Total cost function given in (23.31) is minimized by formulating the
optimization problem
E(q* ,r * ,W * ,S* ,T *) = min E(q,r ,W ,S,T ).

Q<N <M <K ≤L
EI EB
= min{ Ch Ls + (CV 1 + CV 2 + CV 3 ) + C01 1
EC EC
EB2 EB
+ C02 + C03 3 + α Cl P(3,l )}
EC EC
Subject to 1 ≤ q < r < W < S < T ≤ L (23.32)

We use the advances optimization technique known as a Genetic
Algorithm, which is a method based on genetic selection and genetic con-
siderations. In a Genetic Algorithm, there are a bunch of all solutions to
a problem. This solution then deals with regeneration and modification
(as in original genes), produces new a child, and then repeats for differ-
ent generations. Each person is alloted a qualifying value (based on the
value of his or her goal) and the correct and suitable people are given a
chance of getting married and giving more “worthy” people. This is in line
with Darwin’s theory of “Survival of the Fittest”. In this way, we continue to
“change” better people or solutions from generation to generation until we
come to the point of deciding. The steps are as follows:
Getting started: enter value α, β, C, CV1, CV2, CV3, C01, C02, C03 C1, genes (Gq,
Gr, GW, GM, GT), population size, probability and chance of conversion
Output: The standard solution is q *, r * 〖, W〗 *, S * , T] *
Population: It is part of all possible (coded) solutions to a given prob-
lem. The first human population is a fixed G-length size with the genes Gq,
Gr, GW, GM, and GT in the binary of 0 and 1 binary are generated ran-
domly. The first human population that satisfies the inhibitory state (23.32)
of each chromosome is produced randomly.
Calculation Qualification: The genetic value of each chromosome is
calculated.
Choice: According to Darwin’s Theory “Survival of the Fittest”, the right

amount was selected. Here, 3/4 of the population is selected for the next
round according to their qualification value.
Crossover: During crossover, create new people by combining the char-
acteristics of the people we have selected. If young people have a better
fitness level, then the crossover operators will be killed. Here, we consider
the 100% probability of a single point crossover method. Young people
meet the conditions of the challenge.
Conversion: Conversion involves rearranging the list. Genetic modifica-
tion is applied to each child after the crossover. The probability of conver-
sion is 11% and we use the random value generated internally (0, 1). If the
value is below the conversion rate, then it is slightly selected in the random
area and its values are changed.
Stop: Step 4 to Step 7 is repeated until the stop conditions are met. Here,
50 repetitions are considered as stopping conditions. The above steps have
been repeated 35 times.
MATLAB is used to configure programs. Assume the system volume is
L = 20 and the cost of all lost customers is Cl = 50.
Sensitivity analysis is used for whole values based on changes in certain
values.
The model has parameters α and β. The cost elements are as follows:
Case I: Ch = 100, C01 = 100, C02 = 200, C03 = 250, Cv = Cv1+ Cv2+
Cv3 = 300
Case II: Ch = 150, C01 = 100, C02 = 200, C03 = 300, Cv = Cv1+ Cv2+
Cv3 = 300
Case III: Ch = 100, C01 = 150, C02 = 250, C03 = 300, Cv = Cv1+ Cv2+
Cv3 = 500
Case IV: Ch = 100, C01 = 100, C02 = 200, C03 = 300, Cv = Cv1+ Cv2+
Cv3 = 400
Case V: Ch = 100, C01 = 100, C02 = 200, C03 = 300, Cv = Cv1+ Cv2+
Cv3 = 300
Case VI: Ch = 100, C01 = 100, C02 = 250, C03 = 300, Cv = Cv1+ Cv2+
Cv3 = 400
Case VII: Ch = 100, C01 = 100, C02 = 200, C03 = 400, Cv = Cv1+
Cv2+ Cv3 = 200
For all the costs listed above, optimal value and expected cost for dif-
ferent values of α are shown in Table 23.1. It is clearly seen that total cost
increases as α increases. When all the servers are on vacation, Cv increases
(See Case III & Case VII) as it includes the cost associated with Cl and
Table 23.1 Expected cost E (q*, r*, W*, S*, T*) and optimal value (q*, r*, W*, S*, T*) (For α=4, 6, 7, 8, 9, 10).
α 4 6 7 8 9 10
I q*, r*, W*, S*, T* 1,3,4,5,6 2,3,4,5,6 1,2,3,4,6 2,3,4,6,7 4,5,6,7,8 3,4,5,6,7
E(q*, r*, W*, S*, T*) 1099.89 1001.43 1003.90 1005.87 1006.34 1007.21
II q*, r*, W*, S*, T* 2,3,4,5,6 1,2,3,4,6 3,4,5,6,7 2,3,4,5,8 3,4,5,6,8 4,5,6,7,8
E(q*, r*, W*, S*, T*) 1424.34 1461.45 1495.74 1501.21 1503.90 1505.43
III q*, r*, W*, S*, T* 2,3,4,6,7 3,4,5,6,8 1,2,4,6,7 2,4,6,7,8 4,5,7,8,9 5,6,7,8,9
E(q*, r*, W*, S*, T*) 2007.39 2018.14 2100.56 3112.71 3413.20 3740.38
IV q*, r*, W*, S*, T* 3,5,6,7,8 2,4,6,7,8 2,3,4,6,7 4,5,6,7,9 5,6,7,8,9 6,7,8,9,10
E(q*, r*, W*, S*, T*) 1004.13 1007.67 1011.31 1149.74 1179.89 1200.06
V q*, r*, W*, S*, T* 3,4,7,8,9 4,5,6,7,9 2,5,7,8,9 6,7,8,9,10 4,6,8,9,10 5,7,8,9,10
E(q*, r*, W*, S*, T*) 1262.11 1342.03 1376.40 1428.29 1501.10 1503.29
VI q*, r*, W*, S*, T* 4,5,6,7,8 3,4,5,7,9 4,5,7,8,10 5,6,7,9,10 6,7,8,9,10 5,7,8,9,10
E(q*, r*, W*, S*, T*) 1004.52 1007.21 1191.32 1134.69 1187.74 1200.54
VII q*, r*, W*, S*, T* 5,7,8,9,10 6,7,8,9,10 4,6,8,910 5,7,8,9,10 4,5,8,9,10 5,6,7,8,10
E(q*, r*, W*, S*, T*) 1236.00 1332.39 1737.20 2429.31 2519.93 2739.71
holding cost of all failed machines in the system. Cost associated with
higher threshold value is always increased.
23.5 Conclusions
The paper concluded the steady-state results for a multiserver queueing
system with fixed capacity. A cost analysis is done to find the best level (q*,
r*, W*, S*, T*) of breakdown machines in the system using a genetic algo-
rithm. The flexibility of inclusion of servers increases the system efficien-
cies and avoids customer loss due to service delay. Cost analysis at different
system parameters is useful for researchers and engineers.
References
1. Anupama and Solanki, A., “Performance Analysis And Optimization Of
An M/M/3 Queuing System Under Tetrad Policy” International Journal of
Advanced Technology in Engineering and Science (IJATES), ISSN: 2348 –
7550, ISSN (online): 2348 – 7550, Vol.02(11), pp 433-444, November 2014.
2. Bell CE (1971),” Characterization and computation of optimal policies for
operating an M/G/1 queueing system with removable server”. Operations
Research 19:2008–218.
3. Bell CE (1972).,” Optimal operation of an M/G/1 priority queue with remov-
able server”. Operations Research 21:1281–1289.
4. Gnedenko et al. (1969),” Mathematical methods of Reliability Theory”
Academic press, NewYork.
5. Hahn and Sivazlian (1990),” Distribution of the busy period in a control-
lable M\M\2 under the Triadic (0‚K‚N‚M) policy”, J. Appl. Prob., Vol. 27,
pp.425-432.
6. K. H. Wang and B. D. Sivazlian (1989),” Reliability of a system with warm
standbys and repairmen”. Microelectron. Reliab, 29, 849-860.
7. Kuo-Hsiung Wang and Ya-Ling Wang (2002),” Optimal control of an M/M/2
queueing system with finite capacity operating under the triadic (0‚Q‚N‚M) pol-
icy” Math Meth Oper Res 55:447–460.
8. M. N. Gopalan (1975),” Probablistic analysis of single server n-unit system
with (n-1) warm standbys.” Oper Res. 23, 591-595.
9. M. Yadin and P. Naor (1963),” Queueing systems with a removable service
station”. Opl Res. Q. 14, 393-4005.
10. M. J. Magazine (1971),” Optimal control of multi-channel service systems”.
Naval Res. Logist. Q. 177-183.
11. M. J. Sobel (1969),” Optimal average-cost policy for a queue with start-up and
shut-down costs”. Opns Res. 17, 145-162.
12. Rhee H-K, Sivazlian BD (1990),” Distribution of the busy period in a control-
lableM/M/2 queue operating under the triadic (0,K,N,M) policy”. Journal of
Applied Probability 27:425–432.
13. R. Natarajan (1968),” A reliability problem with spares and multiple repair
facilities”. Opns Res. 16, 1041-1057.
14. Subramanian et al. (1976),” Reliability of a repairable system with standbys
failure.” Opns res. 24, 169-176.
15. Wang K-H, Huang H-M (1995),”Optimal control of a removable server in an
M/Ek/1 queueing system with finite capacity”. Microelectronics and Reliability
35:1023–1030.
16. Wang K-H, Ke J-C (2000),” A recursive method to the optimal control of an
M/G/1 queueing system with finite capacity and infinite capacity”. Applied
Mathematical Modelling 24:899–914.
24
An Approach to Ranking of Single
Valued Neutrosophic Fuzzy Numbers
Based on (α, β, γ) Cut Sets
Sunayana Saikia *
Department of Mathematics, Cotton University, Guwahati, Assam, India
Abstract
In most real-world applications, single-valued neutrosophic fuzzy numbers are
employed to make judgments in a fuzzy environment. As a result, single-valued
neutrosophic fuzzy number ranking is crucial. This paper proposes a novel
approach for ranking single-valued neutrosophic fuzzy numbers. The approach
was created by combining the α, β, and γ cuts of neutrosophic fuzzy numbers in a
convex way. In addition, the proposed approach has been used to prove a number
of theorems. This paper also offers a number of numerical examples.
Keywords: Single valued neutrosophic fuzzy numbers, α-cut set, β-cut set,
γ-cut set, ranking of neutrosophic fuzzy numbers, convex combination
24.1 Introduction
Real-life phenomena cannot be classified by real numbers due to the exis-
tence of linguistic concepts. Fuzzy numbers can be used to rate such issues
because fuzzy numbers may deal with verbal words that are ambiguous or
vague. As a result, numerous ways for ranking fuzzy numbers have been
presented at various times. In addition, the ordering of fuzzy numbers has
been thoroughly examined. In the literature, there are several research arti-
cles that describe various strategies for sorting fuzzy numbers. In a nutshell,
no universally accepted method for sorting fuzzy numbers exists. Existing
E-mail: sunayana.saikia@gmail.com
349
methods have been demonstrated to give outcomes that are counter-

intuitive at times. Furthermore, some of these strategies are non-discrimi-
natory and counter intuitive. In 1965, Zadeh [33] “was the first to introduce
the notion of fuzzy set theory”. Many academics have worked hard to come
up with new and better techniques to rank the unclear numbers. Jain [18,
19] introduced the first unorthodox efforts to the condition of “ordering
fuzzy quantities using maximizing set” in 1976. After that, “some rank-
ing methods of fuzzy numbers examined and compared” by Bortolan and
Degani [4] and Burnelli and Mezei [5]. In addition, Wang and Kerre [30,
31], Dubois and Prade [15], Chen [6], Choobineh and Li [9], Yu and Dat
[32], Chen and Sanguansat [7], Chen et al. [8], Shureshjani and Darehmiraki
[25], Chutia [10], Chutia and Chutia [11], Chutia and Gogoi [12], and so
forth proposed some ranking methods and some properties of the ranking
methods. In 1998, Smarandache [26–28] did the first pioneering work on
neutrosophic fuzzy numbers. Ulucay et al. [29], Karaaslan [23], Deli [13],
Jana and Pal [21], Jana et al. [20, 22], and others utilized the neutrosophic
number for a variety of decision-making situations. Several studies have
been conducted on the ordering of single-valued neutrosophic numbers.
Peng et al. [24] established a single-valued neutrosophic number rank-
ing system that was subsequently used to “multi-criteria decision making
issues”. To rank the single-valued neutrosophic numbers, Deli and Subas
[14], Biswas et al. [3], and Aal et al. [1] “presented values and ambiguities of
truth membership, indeterminacy- membership, and falsity- membership
functions”. Garai et al. [34] “devised a single-valued neutrosophic number
ranking system based on possible mean, variance, and standard deviation,
which they subsequently applied to multi-attribute group decision-making
difficulties”. Garai et al. [16] “proposed a weighted possibility means-based
ranking method for single-valued neutrosophic numbers”. Bhaumik et al.
[2] “proposed a new ranking method for single-valued neutrosophic num-
bers established on the (α, β, γ)-cut set, which was then used to the solution
of bi- matrix games”.
Because of their linguistic terms, ranking fuzzy numbers plays a distin-
guishing part in the decision- making process. In some decision-making
scenarios, rating single-valued neutrosophic numbers is crucial. Single-
valued neutrosophic numbers can be ranked using a variety of approaches.
Furthermore, existing single-valued neutrosophic fuzzy number ranking
algorithms are incorrect in several instances. As a result, new ways are
required to overcome present methods’ limits and downsides. This study
makes an effort at doing so. The suggested technique is based on a convex
combination of single-valued neutrosophic numbers from the α, β, and γ
cut sets.
Ranking of Single Valued Neutrosophic Fuzzy Numbers 351
In Section 24.2, some fundamental definitions and arithmetic of single

valued neutrosophic fuzzy numbers are covered. The convex combination
of α-cut, β-cut, and γ-cut sets of a single valued neutrosophic fuzzy num-
ber are proposed in Section 24.3 as a novel ranking technique for single
valued neutrosophic fuzzy numbers. With the aid of the suggested ranking
mechanism, we establish certain theorems in Section 24.4. Some numeri-
cal examples of the suggested technique are presented in Section 24.5. The
proposed method’s conclusion and future studies has been presented in
Section 24.6.
24.2 Definition and Representations

This section looked at several definitions and representations for the idea
of single-valued neutrosophic fuzzy numbers.
Definition 24.2.1 ([11]). “Let A represent a set and X represent the uni-
versal set, with A ∈ X. The set of ordered pairs A = {(x, µA(x)) : x ∈ X, µA :
X → [0, 1]} is then referred to as a fuzzy set and µA is the membership
function”.
Definition 24.2.2 ([11]). “Let a fuzzy number A = (x, µA(x)) = (a, b, c, d;
wA) and if the membership function µA(x) is represented as,
 0, if x < a ,

 x − a  , if a ≤ x ≤ b ,
 b − a 

µ A (x ) =  w A , if b ≤ x ≤ c ,
 d − x 
 , if c ≤ x ≤ d ,
 d − c 
 0, if x > d.

then the fuzzy number A is called a generalized trapezoidal fuzzy number

(GTrFN)”.
Definition 24.2.3 ([11]). “An intuitionistic fuzzy set (IFS) A on  is
defined as A = ⟨x, µA(x), νA(x)⟩ : x ∈ ), where the function µA :  → [0,
1] is known as the degree of membership and νA :  → [0, 1] is known as
the degree of non-membership of the element x ∈ , such that 0 ≤ µA(x)
+ νA(x) ≤ 1”.
Definition 24.2.4 ([11]). “Consider an intuitionistic fuzzy number

(IFN) A = ⟨(a, b, c, d; wA), (p, q, r, s; uA)⟩, where c ≤ a, q ≤ b ≤ c ≤ r, d ≤
s are the parameters. The membership function and the non-membership
function are defined as:
 w A ( x − a)
 b−a , if a ≤ x ≤ b ,

 wA , if b ≤ x ≤ c ,
µ(x ) = 
 w A (d − x ) , if c ≤ x ≤ d ,
 d −c
 0, otherwise.

 (q − x ) + u A (x − p)
 , if p ≤ x ≤ q ,
q− p

 uA , if q ≤ x ≤ r ,
v( x ) = 
 ( x − r ) + u A (s − x ) , if r ≤ x ≤ s ,
 s −r
 1, otherwise.

respectively, where wA is the maximum degree of membership and uA is the

minimum degree of non- membership such that 0 ≤ wA ≤ 1, 0 ≤ uA ≤ 1 and
0 ≤ wA + uA ≤ 1”.
Definition 24.2.5 ([17]). “Assume that X is the universal set and that A
is a set in X. The ordered pair set A = {⟨x, µA(x), νA(x), ρA(x)⟩ : x ∈ X} is a
single-valued neutrosophic set (SVNS), with µA(x) being the truth mem-
bership degree, defined as µA : X → [0, 1], νA(x) being the indeterminacy
membership degree, defined as νA : X → [0, 1] and ρA(x) being the falsity
membership degree is defined as ρA : X → [0, 1], of the element x to the set
A, with the constraint 0 ≤ µA(x) + νA(x) + ρA(x) ≤ 3, ∀x ∈ X”.
Definition 24.2.6 ([17]). “A SVNS A = {⟨x, µA(x), νA(x), ρA(x)⟩∀x ∈ X} is
said to be neutrosophic- normal if there exist at least three points a, b, c ∈
X where µA : a → {1}, νA : b → {1} and ρA : c → {1}”.
Definition 24.2.7 ([17]). “An (α, β, γ)- cut set of a SVNS A is a crisp
subset over the set of real numbers  which are defined as, A(α, β, γ) = {x :
µA(x) ≥ α, νA(x) ≤ β, ρA(x) ≤ γ}, with 0 ≤ α ≤ 1, 0 ≤ β ≤ 1 and 0 ≤ α + β +
γ ≤ 3”.
Definition 24.2.8 ([17]). “A SVNS, A = {⟨x, µA(x), νA(x), ρA(x)⟩∀x ∈ X}

is called neutrosophic-convex if:
(i) µ A(λa + (1 − λ)b) ≥ min(µA(a), µA(b)),
(ii) νA(λa + (1 − λ)b) ≤ max(νA(a), νA(b)),
(iii) ρA(λa + (1 − λ)b) ≤ max(ρA(a), ρA(b)), for all a, b ∈ X and
λ ∈ [0, 1]”.
Definition 24.2.9 ([17]). “A single-valued nutrosophic set {A = ⟨x,

µA(x), νA(x), ρA(x)⟩∀x ∈ X} is said to be a single-valued neutrosophic num-
ber (SVNN) if,
(i) A is neutrosophic-normal,
(ii) A is neutrosophic-convex,
(iii) µA is upper semi-continuous, νA, and ρA are lower semi-
continuous,
(iv) Support (S(A)) of A is bounded, i.e., S(A) = {µA(x) > 0, νA < 1,
ρA < 1, ∀x ∈ X}”.
Definition 24.2.10 ([17]). “A SVNS, A = {⟨x, µA(x), νA(x), ρA(x)⟩∀x ∈
X} = ⟨(a1, b1, c1), (i1, j1, k1), (p1, q1, r1)⟩ in the set of real numbers is said
to be single-valued triangular neutrosophic number (SVTNN), such that
p1 ≤ i1 ≤ a1 ≤ q1 ≤ j1 ≤ b1 ≤ c1 ≤ j1 ≤ r1 are the parameters. Then, the truth
membership degree, the indeterminacy membership degree, and thee fal-
sity membership degree of A can be defined as:”
 x − a1 
 b − a  , if a1 ≤ x ≤ b1 ,
 1 1 
 1, if x = b1 ,
TA (x ) = 
 c1 − x  , iff b1 < x ≤ c1 ,
 c1 − b1 

 0, otherwise.
 j1 − x 
 j − i  , if i1 ≤ x ≤ j1 ,
 1 1 
 0 , if x = j1 ,
I A (x ) = 
 x − j1  , iff j1 ≤ x ≤ k1 ,
 k1 − j1 

 1, otherwise.
 p1 − x 
 q − p  , if p1 ≤ x ≤ q1 ,
 1 1 
 0, if x = q1 ,
FA (x ) = 
 x − j1  , iff j1 ≤ x ≤ k1 ,
 k1 − j1 

 1, otherwise.
Definition 24.2.11 ([2]). Cut Sets of SVTNN

“Let, a SVTNN A = {⟨µA(x), νA(x), ρA(x)⟩ : x ∈ X} = ⟨(a1, b1, c1), (i1, j1, k1),
(p1, q1, r1)⟩, where a1, b1, c1, i1, j1, k1, p1, q1, r1 ∈ R. Then, (α, β, γ)-cut sets of
A are defined as:
(i) -cut set: Aα = {x : µA(x) ≥ α} = [Lα(A), Rα(A)] = [a1 + α(i1 −

α
a1), p1 − α(p1 − i1)],
(ii) β-cut set: Aβ = {x : νA(x) ≤ β} = [Lβ(A), Rβ(A)] = [b1 + β(j1 −
b1), b31 − β(q1 − j1)],
(iii) γ-cut set: Aγ = {x : ρA(x) ≤ γ} = [Lγ(A), Rγ(A)] = [c1 + γ(k1 −
c1), r1 − γ(r1 − k1)]”.
Definition 24.2.12 ([2]). Defined the Quantity Value

“Let A = {⟨µA(x), νA(x), ρA(x)⟩ : x ∈ X} be a SVNN. Then,
1. The value of a SVTNN “A” for the truth membership func-

tion is denoted by Vµ(A) and is defined as
1
∫ f (α )(L (A) + R (A))dα

α α
=
Vµ ( A)
0
where α ∈ [0, 1] and the weighted function f (α) = α satisfies

1
the condition f (0) = 0 and f (1) = 1, such that ∫ f (α )dα =
2.
1
0
2. The value of a SVTNN “A” for the indeterminacy member-
ship function is denoted by Vν(A) and is defined as
∫ g (β )(L (A) + R (A))dβ

β β
=
Vν ( A)
0
where β ∈ [0, 1] and the weighted function g(β) = 1 − β satisfy

1
the condition g(0) = 1 and g(1) = 0, such that ∫ g (β )d β =
2.
1
0
3. The value of a SVTNN “A” for falsity membership function is
denoted by Vρ(A) and is defined as
∫ h(γ )(L (A) + R (A))dγ ,

γ γ
=
Vρ ( A)
0
where γ ∈ [0, 1] and the weighted function h(γ) = 1 − γ satisfy

1
the condition h(0) = 1 and h(1) = 0, such that ∫ 0 h(γ )dγ =
1
2
.
Definition 24.2.13 ([2]). Arithmetic

Let two SVTNNs as A = {⟨µA(x), νA(x), ρA(x)⟩ : x ∈ X} = ⟨(a1, b1, c1), (i1,
j1, k1), (p1, q1, r1)⟩ ∈  and B = {⟨µB(x), νB(x), ρB(x)⟩ : x ∈ X} = ⟨(a2, b2, c2),
(i2, j2, k2), (p2, q2, r2)⟩ ∈ , where µA and µB are truth memberships, νA and
νB are indeterminacy memberships, and ρA and ρB are falsity memberships
of A and B, respectively. Then,
1. “A + B = ⟨(a1 + a2 − a1a2, b1 + b2 − b1b2, c1 + c2 − c1c2), (i1i2, j1j2,

k1k2), (p1p2, p1q2, r1r2)⟩,
2. A × B = ⟨(a1b1, a2b2, c1c2), (i1 + i2 − i1i2, j1 + j2 − j1j2, k1 + k2 −
k1k2), (p1 + p2 − p1p2, q1 + q2 − q1q2, r1 + r2 − r1r2)⟩,
3. κA = ⟨(1 − (1 − a1)κ, 1 − (1 − b1)κ, 1 − (1 − c1)κ),
(i1κ , j1κ ,i1κ ),( p1κ ,q1κ ,r1κ )⟩,
), where κ can be any real number,
4. Aκ = ⟨((a1)κ, (b1)κ, (c1)κ), (1−(1−i1)κ, 1−(1−j1)κ, 1−(1−k1)κ),
((p1)κ, 1−(1−q1)κ, 1−(1−r1)κ)⟩, where κ can be any real number”.
24.3 Proposed Method

Here, a new ranking method for single-valued neutrosophic fuzzy num-
bers (SVNFNs) has been discussed. Consider a fuzzy number A ∈ X to be
a SVTNN, such as A = ⟨x, µA(x), νA(x), ρA(x)⟩ = ⟨[Lα(A), Rα(A)], [Lβ(A),
Rβ(A)], [Lγ(A), Rγ(A)]⟩, 0 ≤ α, β, γ ≤ 1, “where [Lα(A), Rα(A)] is α- cut set,
[Lβ(A), Rβ(A)] is β- cut set and [Lγ(A), Rγ(A)] is γ- cut set for truth mem-
bership function, indeterminacy membership function, and falsity mem-
bership function” of the SVTNN A, respectively.
Then, the proposed ranking index at α, β, γ cuts is defined as
1 1
=
∫
0
α
∫
 λ1 ,λ2 ,λ3 ( A) λ1 f (α )L ( A)dα + (1 − λ1 ) f (α )Rα ( A)dα
0
1 1
∫ ∫
+ λ2 g (β )L ( A)d β + (1 − λ2 ) g (β )R β ( A)d β
β
0 0
1 1
∫ ∫
+ λ3 h(γ )Lγ ( A)dγ + (1 − λ3 ) h(γ )Rγ ( A)dγ
0 0
where λ1, λ2, λ3 ∈ [0, 1].

Thus, the ordering of the fuzzy numbers A, B ∈ X, based on the ranking
index  λ1 ,λ2 ,λ3 is defined as:
a)  λ1 ,λ2 ,λ3 ( A) > λ1 ,λ2 ,λ3 (B), then A ≻ B;
b)  λ1 ,λ2 ,λ3 ( A) <  λ1 ,λ2 ,λ3 (B), then A ≺ B;
c)  λ1 ,λ2 ,λ3 ( A) =  λ1 ,λ2 ,λ3 (B), then A ~ B,
where (0 ≤ λ1, λ2, λ3 ≤ 1).
24.4 Theorems
Some crucial and fundamental theorems are discussed here.
Theorem 24.4.1. Let  λ1 ,λ2 ,λ3 be linear.
Proof. Let, A = ⟨(a1, b1, c1), (i1, j1, k1), (p1, q1, r1)⟩ and B = ⟨(a2, b2, c2), (i2, j2,
k2), (p2, q2, r2)⟩ be two SVTNNs and κ be any real number. Then, it has to
prove that  λ1 ,λ2 ,λ3 ( A + κ B) =
 λ1 ,λ2 ,λ3 ( A) + κ  λ1 ,λ2 ,λ3 (B). Then,
1 1
 λ1 ,λ2 ,λ3 ( A + κ B) is equal to λ1 ∫ f (α )Lα ( A + κ B)dα + (1 − λ1 ) ∫ f (α )
0 0
1 1
Rα ( A + κ B)dα + λ2 ∫ g (β )Lβ ( A + κ B)d β + (1 − λ2 ) ∫ g (β )R β ( A + κ B)d β + λ3
0 0
1 1 1
∫ h(γ )Lγ ( A + κ B)dγ + (1 − λ3 ) ∫ h(γ )Rγ ( A + κ B)dγ ,which leads to λ1 ∫
0 0 0
1 1
f (α )(Lα ( A) + κ Lα (B))dα + (1 − λ1 ) ∫ f (α )(Rα ( A) + κ Rα (B))dα + λ2 ∫ g (β )
0 0
1 1
(Lβ ( A) + κ Lβ (B))d β + (1 − λ2 ) ∫ g (β )(R β ( A) + κ R β (B))d β + λ3 ∫ h(γ )(Lγ ( A) +
0 0
1 1
κ Lγ (B))dγ + (1 − λ3 ) ∫ h(γ )(Rγ ( A) + κ Rγ (B))dγ , then it is true that(λ1 ∫ f (α )
0 0
1 1 1
Lα ( A)dα + (1 − λ1 ) ∫ f (α )Rα ( A)dα + λ2 ∫ g (β )Lβ ( A)d β + (1 − λ2 ) ∫ g (β )R β
0 0 0
1 1 1
( A)d β + λ3 ∫ h(γ )Lγ ( A)dγ + (1 − λ3 ) ∫ h(γ )Rγ ( A)dγ ) + κ (λ1 ∫ f (α )Lα (B)dα +
0 0 0
3 1
0 0 0
1 1
f (α )(Lα ( A) + κ Lα (B))dα + (1 − λ1 ) ∫ f (α )(Rα ( A) + κ Rα (B))dα + λ2 ∫ g (β )
0 0
1 1
(Lβ ( A) + κ Lβ (B))d β + (1 − λ2 ) ∫ g (β )(R β ( A) + κ R β (B))d β + λ3 ∫ h(γ )(Lγ ( A) +
0 0
Ranking of Single
1
Valued Neutrosophic Fuzzy Numbers 1 357
κ Lγ (B))dγ + (1 − λ3 ) ∫ h(γ )(Rγ ( A) + κ Rγ (B))dγ , then it is true that(λ1 ∫ f (α )
0 0
1 1 1
Lα ( A)dα + (1 − λ1 ) ∫ f (α )Rα ( A)dα + λ2 ∫ g (β )Lβ ( A)d β + (1 − λ2 ) ∫ g (β )R β
0 0 0
1 1 1
( A)d β + λ3 ∫ h(γ )Lγ ( A)dγ + (1 − λ3 ) ∫ h(γ )Rγ ( A)dγ ) + κ (λ1 ∫ f (α )Lα (B)dα +
0 0 0
1 1 1
(1 − λ1 ) ∫ f (α )Rα (B)dα + λ2 ∫ g (β )Lβ (B)d β + (1 − λ2 ) ∫ g (β )R β (B)d β
0 0 0
1 1
+λ3 ∫ h(γ )Lγ (B)dγ + (1 − λ3 ) ∫ h(γ )Rγ (B)dγ ), then the result follows
0 0
that  λ1 ,λ2 ,λ3 ( A) + k  λ1 ,λ2 ,λ3 (B).
Theorem 24.4.2. If A, B, and C are three arbitrary SVTNNs and A ≲ B and

B ≲ C are true, then A ≲ C is true.
Proof. Given that A ≲ B, then  λ1 ,λ2 ,λ3 ( A) ≤  λ1 ,λ2 ,λ3 (B), for 0 ≤ λ1, λ2,
λ3 ≤ 1. Also, given that B ≲ C, then  λ1 ,λ2 ,λ3 (B) ≤  λ1 ,λ2 ,λ3 (C ), for 0 ≤ λ1,
λ2, λ3 ≤ 1. This implies that  λ1 ,λ2 ,λ3 ( A) ≤  λ1 ,λ2 ,λ3 (C ), for 0 ≤ λ1, λ2, λ3 ≤ 1.
Hence, the result follows that A ≲ C.
24.5 Numerical Examples

In this section, some numerical examples are discussed to assess how well
the proposed method performs.
Example 24.5.1. Consider two SVTNNs A = ⟨(0.50, 0.65, 0.80), (0.10,
0.15, 0.30), (0.10, 0.20, 0.30)⟩, B = ⟨(0.10, 0.20, 0.30), (0.20, 0.30, 0.40),
(0.40, 0.50, 0.70)⟩ and three weighted functions be P = (λ1 = 0.3, λ2 = 0.4,
λ3 = 0.5), Q = (λ1 = 0.5, λ2 = 0.5, λ3 = 0.5) and S = (λ1 = 1.0, λ2 = 1.0, λ3 = 0.9).
Then, the comparison is between two SVTNNs, A and B, as shown in
Table 24.1.
Example 24.5.2. Consider two SVTNNs A = ⟨(0.30, 0.45, 0.50), (0.10,
0.20, 0.40), (0.10, 0.20, 0.30)⟩, B = ⟨(0.20, 0.30, 0.35), (0.10, 0.10, 0.10),
(0.60, 0.70, 0.80)⟩ and three weighted functions P = (λ1 = 1.0, λ2 = 1.0, λ3 =
1.0), Q = (λ1 = 0.5, λ2 = 0.5, λ3 = 0.5) and S = (λ1 = 0.1, λ2 = 0.1, λ3 = 0.1).
Then, the comparison is between the two SVTNNs, A and B, as shown
in Table 24.2.
Example 24.5.3. Consider two SVTNNs A = ⟨(0.6, 0.7, 0.9), (0.1, 0.3,
0.4), (0.1, 0.4, 0.6)⟩, B = ⟨(0.1, 0.2, 0.3), (0.4, 0.5, 0.6), (0.7, 0.8, 0.9)⟩ and
three weighted functions P = (λ1 = 0.3, λ2 = 0.4, λ3 = 0.5), Q = (λ1 = 0.3, λ2 =
0.6, λ3 = 0.5) and S = (λ1 = 0.6, λ2 = 0.4, λ3 = 0.7). Then, the comparison is
between the two SVTNNs, A and B, as shown in Table 24.3.
Table 24.1 Ranking of SVNNs in Example 24.5.1.
λ1,λ2,λ3  λ1 , λ2 , λ3 (A)  λ1 , λ2 , λ3 (B) Results

0.3,0.4,0.5 0.5217 0.5384 A≺B
0.5,0.5,0.5 0.5083 0.5083 A ~B
1.0,1.0,0.9 0.4534 0.45502 A≺ B

1.0,1.0,1.0 0.0669 0.5168 A≺ B
0.5,0.5,0.5 0.4251 0.5460 A≺ B
0.1,0.1,0.1 0.4719 0.5693 A≺ B

0.3,0.4,0.5 0.6584 0.8101 A≺ B
0.3,0.6,0.5 0.7550 0.7500 A≻B
0.6,0.4,0.7 0.7000 0.7401 A≺ B
24.6 Conclusion
The majority of real-life decision-making is based on fuzzy numbers due
to their language connotations. Ordering fuzzy numbers is much more
important in order to solve such difficulties. There are several documented
ranking systems for single-valued fuzzy numbers. This paper developed a
novel ranking approach for single-valued fuzzy numbers based on convex
combinations of the fuzzy numbers’ α, β, and γ cut sets. A number of the-
orems have also been proved using the proposed method. Furthermore,
numerical examples have been provided to demonstrate the suggested
strategy’s superior performance. This ranking method may be used to look
at the rankings of intuitionistic fuzzy numbers, type-2 fuzzy numbers,
hesitant fuzzy numbers, and other fuzzy numbers. Further, this proposed
method can be applied to real-life problems like medicine, business, eco-

nomics, computer sciences, etc.
References
1. S. I. A. Aal, M. M. A. A. Ellatif, and M. M. Hassan. Two Ranking Methods
of Single Valued Triangular Neutrosophic Numbers to Rank and Evaluate
Information Systems Quality, 2018.
2. A. Bhaumik, S. Roy, and D.-F. Li. (α, β, γ)-cut set based ranking approach
to solving bi-matrix games in neutrosophic environment. Soft Computing,
25:1–11, 02 2021.
3. P. Biswas, S. Pramanik, and B. Giri. Value and ambiguity index based rank-
ing method of single-valued trapezoidal neutrosophic numbers and its appli-
cation to multi-attribute decision making. Neutrosophic Sets and Systems,
12:127–138, 07 2016.
4. G. Bortolan and R. Degani. A review of some methods for ranking fuzzy
subsets. Fuzzy Sets and Systems, 15(1):1 – 19, 1985.
5. M. Brunelli and J. Mezei. How different are ranking methods for fuzzy num-
bers? a numerical study. International Journal of Approximate Reasoning,
54(5):627 – 639, 2013.
6. S. M. Chen. New methods for subjective mental workload assessment and
fuzzy risk analysis. Cybernetics and Systems, 27(5):449–472, 1996.
7. S.-M. Chen and K. Sanguansat. Analyzing fuzzy risk based on a new fuzzy
ranking method between generalized fuzzy numbers. Expert Systems with
Applications, 38(3):2163 – 2171, 2011.
8. Z. Chen, G. Huang, and A. Chakma. Hybrid fuzzy-stochastic modeling
approach for assessing environmental risks at contaminated groundwater
systems. Journal of Environmental Engineering, 129(1):79–88, 2003.
9. F. Choobineh and H. Li. An index for ordering fuzzy numbers. Fuzzy Sets
and Systems, 54(3):287 – 294, 1993.
10. R. Chutia. Ranking of fuzzy numbers by using value and angle in the epsi-
lon-deviation degree method. Applied Soft Computing, 60:706 – 721, 2017.
11. R. Chutia and B. Chutia. A new method of ranking parametric form of fuzzy
numbers using value and ambiguity. Applied Soft Computing, 52:1154 – 1168,
2017.
12. R. Chutia, R. Gogoi, and D. Datta. Ranking p-norm generalised fuzzy
numbers with different left height and right height using integral values.
Mathematical Sciences, 9(1):1–9, 2015.
13. I. Deli. Operators on single valued trapezoidal neutrosophic numbers and
svtn-group decision making. Neutrosophic Sets and Systems, 22:131–150.
14. I. Deli and Y. uba. A ranking method of single valued neutrosophic num-
bers and its applications to multi-attribute decision making problems.
International Journal of Machine Learning and Cybernetics, 8, 08 2017.
15. D. Dubois and H. Prade. Fuzzy Sets and Systems: Theory and Applications.
Academic Press, Inc., Orlando, FL, USA, 1980.
16. T. Garai, H. Garg, and T. Roy. A ranking method based on possibility mean
for multi-attribute decision making with single valued neutrosophic num-
bers. Journal of Ambient Intelligence and Humanized Computing, 11, 11 2020.
17. P. Grzegorzewski. Algorithms for trapezoidal approximations of fuzzy num-
bers preserving the expected interval. Fuzzy Sets and Systems, 159:1354–
1364, 06 2008.
18. R. Jain. Decisionmaking in the presence of fuzzy variables. IEEE Transactions
on Systems, Man and Cybernetics, SMC-6(10):698–703, Oct 1976.
19. R. Jain. A procedure for multiple-aspect decision making using fuzzy sets.
International Journal of Systems Science, 8(1):1–7, 1977.
20. C. Jana, G. Muhiuddin, and M. Pal. Multiple-attribute decision-making
problems based on svtnh methods. Journal of Ambient Intelligence and
Humanized Computing, 11, 09 2020.
21. C. Jana and M. Pal. A robust single-valued neutrosophic soft aggregation
operators in multi-criteria decision making. Symmetry, 11(1), 2019.
22. C. Jana, M. Pal, F. Karaaslan, and J. Wang. Trapezoidal neutrosophic aggrega-
tion operators and their application to the multi-attribute decision-making
process. Scientia Iranica, 27(3):1655–1673, 2020.
23. F. KARAASLAN. Gaussian single-valued neutrosophic numbers and its appli-
cation in multi- attribute decision making. Infinite Study, 2018.
24. J. Peng, J. Wang, H.-Y. Zhang, and X.-h. Chen. An outranking approach for
multi-criteria decision-making problems with simplified neutrosophic sets.
Applied Soft Computing, 25:336–346, 12 2014.
25. R. A. Shureshjani and M. Darehmiraki. A new parametric method for rank-
ing fuzzy numbers. Indagationes Mathematicae, 24(3):518 – 529, 2013.
26. F. Smarandache. Neutrosophy: Neutrosophic Probability, Set, and Logic:
Analytic Synthesis & Synthetic Analysis. Analytic Synthesis and Synthetic
Analysis Series. American Research Press, 1998.
27. F. Smarandache. A unifying field in logics: Neutrosophic logic., 1999.
28. F. Smarandache. Neutrosophic set a generalization of the intuitionistic fuzzy
set. International Journal of Pure and Applied Mathematics, 24:38–42, 01
2004.
29. V. Ulucay, A. Kilic, I. Yildiz, and M. ahin. A new approach for multi-attribute
decision-making problems in bipolar neutrosophic sets. Infinite Study, 2018.
30. X. Wang and E. E. Kerre. Reasonable properties for the ordering of fuzzy
quantities (i). Fuzzy Sets and Systems, 118(3):375 – 385, 2001.
31. X. Wang and E. E. Kerre. Reasonable properties for the ordering of fuzzy
quantities (ii). Fuzzy Sets and Systems, 118(3):387 – 405, 2001.
32. V. F. Yu and L. Q. Dat. An improved ranking method for fuzzy numbers with
integral values. Applied Soft Computing, 14, Part C:603 – 608, 2014.
33. L. A. Zadeh. Fuzzy sets. Information and Control, 8(3):338–353, 1965.
34. H.-Y. Zhang, J. Wang, and X. Chen. An outranking approach for multi-
criteria decision-making problems with interval-valued neutrosophic sets.
Neural Computing and Applications, 27, 04 2015.
25
Performance Analysis of Database Models
Based on Fuzzy and Vague Sets for
Uncertain Query Processing
Sharmistha Ghosh1* and Surath Roy2
Department of Basic Science & Humanities, Institute of Engineering &

1
Management, Kolkata, West Bengal, India

2
Department of Mathematics, Brainware University, Kolkata, West Bengal, India
Abstract
One of the primary aspects of utilization of any database model lies in its potential
in processing information and queries accurately. In the present work, the authors
intend to make a comparative analysis on the capability of fuzzy and vague rela-
tional database models in treating uncertain queries. A new algorithm is proposed
and query testing related to a real life example is performed. The investigation
demonstrates that a relational data model based on vague set theory produces
more refined decisions than a fuzzy data model. It may thus be asserted that a
relational database management system (RDBMS) using vague theoretic concept
might lead to better software fabrication than the presently accessible ones.
Keywords: Fuzzy set, vague set, database model, similarity measure, SQL,
fuzzy SQL, vague SQL
25.1 Introduction
In real life, information is very often imprecise or incomplete. The conven-
tional relational database system fails to treat such uncertain data. The the-
ory of fuzzy sets, as formalized by Zadeh [20] in 1965, is extensively applied
to handle such inexact or imprecise data. Several authors [1, 6, 10, 11, 14, 15,
17–19] have also worked on formulation of a query language for a database
representation based on fuzzy sets. Bosc et al. [1] and Nakajima et al. [15]
*Corresponding author: sharmisthag71@gmail.com
363
have outstretched the familiar SQL language in the framework of fuzzy set
theory, namely, SQLF. More recently, Moreau et al. [14] have discussed a
procedure wherein one may fetch data with no prior information about the
database constitution or precise query language. The potential benefits and
efficacy of the use of fuzzy query in a traditional data model have been thor-
oughly explained in [17]. An intelligent approach to extend SQL language to
treat flexible conditions in queries was proposed by Mama et al. [11] in 2021.
Gau and Buehrer [4] initiated the theory of vague sets in 1993 as an
abstraction of a fuzzy set theoretic approach. It is believed that a vague set
that employs interval-based membership values may process uncertain
information in a more efficient manner compared to a fuzzy set. The vague
theoretic concept was further embodied in relations by Lu and Ng [8, 9] and
a new query language called VSQL evolved. A vague database model was
designed by Zhao and Ma [21] and vague querying strategies with SQL were
also investigated. In [3], Dutta et al. have used vague sets to rejuvenate “vague
search”, a new method of intelligent search that is capable of answering any
uncertain query put forth by the user. In [12], the detailed design aspects of
a vague database model have been thoroughly discussed. An architecture for
processing of hesitant queries has been designed by Mishra et al. [13] along
with a comparison of fuzzy and vague sets in handling imprecise queries.
In the present work, the authors also aspire to analyze the ability of fuzzy
and vague database models in relation to uncertain query processing. It has
been noticed that the algorithms used in literature (see refs. [10, 13, 16]) to
obtain membership values possess certain drawbacks. In the current inves-
tigation, the authors present a novel algorithm which is devoid of such defi-
ciencies. The proposed algorithm generates membership functions for fuzzy
or vague sets for the calculation of membership values that are independent
of the attribute. However, it is worthwhile to mention that the membership
functions employed earlier in literature [10, 16] depend on the attribute type.
The framework of the chapter is as follows. The definitions of fuzzy
and vague sets as well as related fundamental concepts are presented in
Sections 25.2.1 and 25.2.2. The similarity measure formulae used in this
study appear in Section 25.2.3. The algorithm designed in the present anal-
ysis to generate membership values is proposed in Section 25.3. Real life
examples are demonstrated in Section 25.4 to observe that a vague theo-
retic approach is better suited in processing queries that are not precise.
The concluding remarks are reported in Section 25.5.
25.2 Basic Definitions

Let X denote the universe of discourse and x represent an element of X.
Performance Analysis of Database Models 365
25.2.1 Fuzzy Set

Definition 25.2.1.1: A fuzzy set S, defined in the universe of discourse X, is
a set of ordered pairs S = {( x , µ S ( x )) : x ∈ X } , where µS : X → [0,1] denotes
the grade of membership of x in S.
It may be easily observed that an ordinary subset A of X may be treated
as a fuzzy set with membership function µA that takes binary values, i.e.,
 1 if x ∈A
µA = 
0 if x ∉A

25.2.2 Vague Set

Definition 25.2.2.1: A vague set S, in the universe of discourse X, is char-
acterized by two membership functions, namely:
(i) a truth function tS : X → [0,1] and

(ii) a false function fS : X → [0,1].
Here, tS(x) represents a lower bound on the grade of membership of x

as obtained from the ‘evidence in favor of x’, whereas fS(x) denotes a lower
bound on the negation of x as deduced from the ‘evidence against x’ and
tS(x) + fS(x) ≤ 1.
The grade of membership µS(x) of x in the vague set S is then bounded
by a subinterval [t S (x ), 1 − f s (x )] of [0, 1], i.e., tS(x) ≤ µS(x) ≤ 1 – fS(x).
Now, the vague set S may be represented = as S {
x ,[t S (x ), 1 − f S (x )] :
: x ∈ X } . The interval [t S (x ), 1 − f s (x )] is termed as the vague value of the
element x. It may be noted that if tS(x) is equal to (1 − f S (x ) ) , the informa-
tion about object x is precise and the theory of vague set degenerates to that
of a fuzzy set. In case both tS(x) and (1 − f S (x ) ) are 1, which confirms that x
belongs to S, the information about x is exact, and the theory reverts to that
of an ordinary set. Similarly, if tS(x) and (1 − f S (x ) ) both take the value 0,
the knowledge about x is again exact which relates to the situation that x in
not in S. Hence, a crisp set as well as fuzzy set can be contemplated to be a
specific case of vague set.
25.2.3 Similarity Measure

The concept of similarity of vague sets has been studied by several research-
ers [2, 5, 7, 8]. The similarity measure proposed by Lu et al. in [8] was
shown to be more effective in general cases. This measure has been used in
this work, which is presented below:
Definition 25.2.3.1: Similarity Measure for Vague Values

Let u and v be vague values such that= u [t1 , 1 − f1 ] and=v [t 2 , 1 − f 2 ] ,
where 0 ≤ t1 ≤ 1 – f1 ≤ 1 and 0 ≤ t2 ≤ 1 – f2 ≤ 1. If SM(u, v) denotes the sim-
ilarity measure between u and v, then
 (t1 − t2 ) − ( f1 − f2 ) 
SM (u,=
v) 1 −
 2 
(
 1 − ( t1 − t 2 ) + ( f1 − f 2 ) . )
Definition 25.2.3.2: Similarity Measure for Vague Sets
Let X = {x1 , x2 , x3, ..., xn } be the universe. Let S1 and S2 be two vague sets
of X, such that
=S1 { x ,[t
i S1 (xi ),1 − f S1 (xi )] : xi ∈ X} ,
where t S1 ( xi ) ≤ µ S1 ( xi ) ≤ 1 − f S1 ( xi ) and 1 ≤ i ≤ n
and
S2 = { xi ,[t S2 ( xi ),1 − f S2 ( xi )] : xi ∈ X } ,
where t S2 ( xi ) ≤ µ S2 ( xi ) ≤ 1 − f S2 ( xi ) and 1 ≤ i ≤ n.
Now, the similarity measure between S1 and S2, denoted by SM (S1 , S2 ) ,
is defined as:
SM ( S1 , S2 ) =
1
n ∑ SM ([t
i =1
S1 ( xi ),1 − f S1 ( xi )],[t S2 ( xi ),1 − f S2 ( xi )])
25.3 Algorithm to Generate Membership Values

In the literature, various membership functions have been employed by dif-
ferent authors for finding the membership values for different fuzzy attributes.
For an EMPLOYEE database, the membership function deployed by
Raju et al. [16] corresponding to a fuzzy set ‘close to u’ and fuzzy attribute
Salary is the following:
1
µclose to u ( x ) = .
|x − u|
1+
20000
However, for the Experience attribute, Raju et al. defined the member-
ship function as
1
µclose to u ( x ) = .
|x − u|
1+
4
On the contrary, Ma et al. [10] used the following membership func-

tions for the same fuzzy attributes, Salary and Experience:
1
µclose to u ( x ) = 2 ;
 |x − u| 
1+
 20000 
1
µclose to u ( x ) = 2.
 |x − u|
1+
 4 
It is obvious that the above membership functions depend on the spe-

cific attribute under investigation.
In this study, the authors have made an effort to devise an algorithm for
calculation of membership values for different fuzzy/vague attributes. It
may be pointed out that the membership function generated by said algo-
rithm is attributed as independent for a fuzzy/vague set. The membership
values generated by the proposed algorithm also compare well with that of
others obtained in the literature.
A similar algorithm was designed by Mishra et al. [13] for the purpose
of estimation of membership values. It is observed that the same algorithm
has been utilized by Yadav in [19] for SQL query processing. But, the for-
mula put forth in [13] is found to possess the following deficiencies:
(i) Negative membership value is generated in certain instances.

(ii) Addition of new records in the given data leads to change of
membership values for the existing data that is not sensible.
The algorithm devised in this study is presented below. It may be noted

that it is devoid of such anomalies.
Algorithm 25.3.1
Input: Fuzzy or vague data and attributes in an imprecise query
Output: Membership value lying in [0,1]
Method:
Fetch attributes from the query
for each fuzzy or vague attribute do
begin
f_data = data value of the attribute from the query
range = maximum domain value – minimum domain value
for every tuple in the relation do
begin
t_value = tuple value from the domain of fuzzy or vague attribute
if (|f_data – t_value | ≤ range)
|f _ data − t _ value|
membership _ value = 1 −
range
else membership_value = 0
end for loop for tuple
end for loop for fuzzy or vague attribute
25.4 Real Life Applications

We now use the following EMPLOYEE relation (Table 25.1) as a real life
example and process certain imprecise queries with one or more attributes.
A comparative analysis is presented in this section for each of the queries
on the framework of fuzzy and vague models.
Query 1: “To fetch the employees with age close to 53 years”.

Query 2: “To retrieve the data of the employees whose age is more
or less than 53 years with work experience close to 22 years.”
The fuzzy and vague database models were used to execute the above
queries with one or more attributes. In each case, the vague database model
produced better results than the fuzzy data model, as may be observed
from the analysis presented herein.
Query 1: “To fetch the employees with age close to 53 years”.

Table 25.1 EMPLOYEE relation.

Name Age (years) Experience (years) Remuneration (₹)
Mr. Roy 30 4 32000
Mr. Roychowdhury 31 6 32500
Mr. Pramanik 34 8 34000
Mr. Barik 51 18 58000
Mr. Bose 57 22 77900
Mr. Ghosh 49 17 57400
Mr. Samanta 50 16 56800
Mr. Bhowmick 41 13 46500
Mr. Banerjee 39 12 42000
Mr. Chattopadhyay 53 20 80900
Mr. Ganguly 52 19 80200
Mr. Kher 56 21 81100
Mr. Mukhopadhyay 54 23 82200
Mr. Patnaik 55 27 82000
Mr. Sahoo 62 35 122000
Solution with Fuzzy Model:

The fuzzy characteristic here is Age and the fuzzy data is close to 53 years.
We employ the algorithm discussed in Section 25.3 to find the member-
ship value for each domain value of the Age attribute.
Here,
Domain of Age attribute = {30, 31, 34, 51, 57, 49, 50, 41, 39, 53, 52, 56,
54, 55, 62}
f_data = 53
Range = 62 – 30 = 32
We now deploy the formula provided in Algorithm 25.3.1 to determine
the membership values for every domain value of Age attribute as follows:
1st tuple: Membership Value = 1 - (|53–30| / 32) = 0.28125

2nd tuple: Membership Value = 1 - (|53–31| / 32) = 0.3125
The complete record of membership values for each tuple is shown in

Table 25.2.
Table 25.3 displays the complete representation of this relation corre-
sponding to Query 1. For the fuzzy attribute Age, the fuzzy representation
appears in column 3 and its equivalent vague formulation occurs in col-
umn 4. This vague formulation is now utilized to compute the similarity
measures (SM) with the fuzzy data ‘close to 53’, which may be represented
as <53, [1, 1]> in vague notation. The formula for similarity measure, as
introduced in Definition 25.2.3.1, is then applied.
If one considers the vague data u = <53,[1,1]> and v = <30,[0.28125,
0.28125]>, then
t1 = 1, f1 = 0, t2 = 0.28125, f2 = 0.71875
Table 25.2 Membership values.

Name Age (years) Membership value
Mr. Roy 30 0.28125
Mr. Roychowdhury 31 0.3125
Mr. Pramanik 34 0.40625
Mr. Barik 51 0.9375
Mr. Bose 57 0.875
Mr. Ghosh 49 0.875
Mr. Samanta 50 0.90625
Mr. Bhowmick 41 0.625
Mr. Banerjee 39 0.5625
Mr. Chattopadhyay 53 1.0
Mr. Ganguly 52 0.96875
Mr. Kher 56 0.90625
Mr. Mukhopadhyay 54 0.96875
Mr. Patnaik 55 0.9375
Mr. Sahoo 62 0.71875
Table 25.3 Similarity measures for Query 1 with Fuzzy Model.

Fuzzy Vague
Age representation representation SM of age with Experience Remuneration
Name (years) of age of age <53,[1,1]> (years) (₹) SM (tuple)
Mr. Roy 30 <30, 0.28125> <30, [0.28125, 0.53033 4 32000 0.53033
0.28125]>
Mr. Roychowdhury 31 <31, 0.3125> <31, [0.3125, 0.55902 6 32500 0.55902

0.3125]>
Mr. Pramanik 34 <34, 0.40625> <34, [0.40625, 0.63738 8 34000 0.63738

0.40625]>
Mr. Barik 51 <51, 0.9375> <51, [0.9375, 0.96825 18 58000 0.96825

0.9375]>
Mr. Bose 57 <57, 0.875> <57, [0.875, 0.875]> 0.93541 22 77900 0.93541
Mr. Ghosh 49 <49, 0.875> <49, [0.875, 0.875]> 0.93541 17 57400 0.93541
Mr. Samanta 50 <50, 0.90625> <50, [0.90625, 0.95197 16 56800 0.95197

0.90625]>
Mr. Bhowmick 41 <41, 0.625> <41, [0.625,. 0.625]> 0.79057 13 46500 0.79057
Mr. Banerjee 39 <39, 0.5625 > <39, [0.5625, 0.75000 12 42000 0.75000
0.5625]>
Mr. Chattopadhyay 53 <53, 1> <53, [1,1]> 1 20 80900 1

(Continued)
Table 25.3 Similarity measures for Query 1 with Fuzzy Model. (Continued)
Fuzzy Vague
Age representation representation SM of age with Experience Remuneration
Name (years) of age of age <53,[1,1]> (years) (₹) SM (tuple)
Mr. Ganguly 52 <52, 0.96875> <52, [0.96875, 0.98425 19 80200 0.98425
0.96875]>
Mr. Kher 56 <56, 0.90625 > <56, [0.90625, 0.95197 21 81100 0.95197
0.90625]>
Mr. Mukhopadhyay 54 <54, 0.96875 > <54, [0.96875, 0.9842 23 82200 0.9842
0.96875]>
Mr. Patnaik 55 <55, 0.9375> <55, [0.9375, 0.96825 27 82000 0.96825

0.9375]>
Mr. Sahoo 62 <62, 0.71875> <62, [0.71875, 0.84779 35 122000 0.84779

0.71875]>
and thus,
SM (u, v )
 (1 − 0.28125) − (0 − 0.71875) 
= 1 −  (1 − (1 − 0.28125) + (0 − 0..71875) )
 2 
= 1 −=
0.71875 =
0.28125 0.53033
Again, for u = <53,[1,1]> and v = <31,[0.3125,0.3125]>,
t1 = 1, f1 = 0, t2 = 0.3125, f2 = 0.6875.
Then,
SM (u, v )
 (1 − 0.3125) − (0 − 0.6875) 
= 1 −  (1 − (1 − 0.3125) + (0 − 0.68775) )
 2 
= 1 −=
0.6875 =
0.3125 0.55902
and so on. The complete results are shown in Table 25.3.

Now, the following SQL statement is generated to execute the given
query at an α-cut or threshold value 0.95, provided by the decision maker:
SELECT * FROM EMPLOYEE WHERE SM(tuple) ≥ 0.95.
Table 25.4 now presents the resultant tuples retrieved from the EMPLOYEE
database.
Table 25.4 Resulting tuples for Query 1 with Fuzzy Model at α= 0.95.
Mr. Barik 51 18 58000
Mr. Samanta 50 16 56800
Mr. Ganguly 52 19 80200
Mr. Kher 56 21 81100
Mr. Patnaik 55 27 82000
Table 25.5 Similarity measures for Query 1 with Vague Model.

Age Vague representation SM of age with Experience Remuneration
Name (years) of age <53,[1,1]> (years) (₹) SM (tuple)
Mr. Roy 30 <30, [0.28125, 0.440416] 0.537289 4 32000 0.537289
Mr. Roychowdhury 31 <31, [0.3125, 0.485216]> 0.574358 6 32500 0.574358
Mr. Pramanik 34 <34, [0.40625, 0.516927]> 0.640691 8 34000 0.640691
Mr. Barik 51 <51, [0.9375, 0.953667]> 0.964182 18 58000 0.964182
Mr. Bose 57 <57, [0.875, 0.878291]> 0.929415 22 77900 0.929415
Mr. Ghosh 49 <49, [0.875, 0.875770]> 0.930180 17 57400 0.930180
Mr. Samanta 50 <50, [0.90625,0.921369]> 0.951472 16 56800 0.951472
Mr. Bhowmick 41 <41, [0.625, 0.696428]> 0.785632 13 46500 0.785632
Mr. Banerjee 39 <39, [0.5625, 0.681963]> 0.743212 12 42000 0.743212
Mr. Chattopadhyay 53 <53, [1.000000,1.00000]> 1.000000 20 80900 1.000000
Mr. Ganguly 52 <52, [0.96875, 0.977372]> 0.982317 19 80200 0.982317
Mr. Kher 56 <56, [0.90625, 0.914506]> 0.946656 21 81100 0.946656
Mr. Mukhopadhyay 54 <54, [0.96875,0.976538]> 0.982701 23 82200 0.982701
Mr. Patnaik 55 <55, [0.9375, 0.943755]> 0.966286 27 82000 0.966286
Mr. Sahoo 62 <62, [0.71875, 0.783282]> 0.835870 35 122000 0.835870
Solution with Vague Model:

We now analyze the same query using a vague data model. Age is the vague
attribute and vague data is ‘close to 53’.
To determine the vague representation of the Age attribute, Algorithm
25.3.1 is now utilized to obtain the truth membership values. On the other
hand, the decision maker provides the false membership values randomly
along with the constraint that the sum of the two membership values can-
not exceed 1. Next, the similarity measures are computed and are displayed
in Table 25.5.
The tuples recovered from the EMPLOYEE relation at this α-cut value
0.95 are shown in Table 25.6.
It may be clearly observed from Tables 25.4 and 25.6 that the vague
model is better off than its fuzzy version. The tuple of Mr. Kher, aged 56,
was not retrieved by the SQL statement with vague query although it was
fetched as fuzzy query. It is worthwhile to note that compared to the other
values returned by the SQL statement, 56 is less close to 53.
We now look at another imprecise question that has two fuzzy or vague
characteristics.
Query 2: “To retrieve the data of the employees whose age is more
or less than 53 years with work experience close to 22 years.”
Solution with Fuzzy Model:

The current query has two attributes: Age and Experience. Table 25.7
shows the fuzzy representation of the EMPLOYEE relationship produced
using Definition 25.2.3.1 and Algorithm 25.3.1.
Table 25.6 Resulting tuples for Query 1 with Vague Model at α = 0.95.
Mr. Barik 51 18 58000
Mr. Samanta 50 16 56800
Mr. Ganguly 52 19 80200
Mr. Patnaik 55 27 82000
Table 25.7 Similarity measures for Query 2 with Fuzzy Model.

Vague Fuzzy SM of
Fuzzy represen represen SM of age represen Vague represen experience
Age tation of tation of with <53, Experience tation of tation of with Remuneration SM
Name (years) age age [1,1]> (years) experience experience <22,1,1]> (₹) (tuple)
Mr. Roy 30 <30, 0.28125> <30, [0.28125, 0.53033 4 <4,0.419355> <4,[0.4193550 0.647576 32000 0.53033
0.28125]> .419355]>
Mr. Roychowdhury 31 <31, 0.3125> <31, [0.3125, 0.55902 6 <6,0.483871> <6,[0.483871, 0.695608 32500 0.55902
0.3125]> 0.483871]>
Mr. Pramanik 34 <34, 0.40625> <34, [0.40625, 0.63738 8 <8, 0.548387> <8,[ 0.548387, 0.740532 34000 0.63738
0.40625]> 0.548387]>
Mr. Barik 51 <51, 0.9375> <51, [0.9375, 0.96825 18 <18,.0.870968> <18,[0.870968, 0.933257 58000 0.933257
0.9375]> 0.870968]>
Mr. Bose 57 <57, 0.875> <57, [0.875, 0.93541 22 <22,1> <22,[1,1]> 1.000000 77900 0.93541
0.875]>
Mr. Ghosh 49 <49, 0.875> <49, [0.875,. 0.93541 17 <17,0.838710> <17,[0.838710, 0.915811 57400 0.915811
0.875]> 0.838710]>
Mr. Samanta 50 <50, 0.90625> <50, [0.90625, 0.95197 16 <16,0.806452> <16,[0.806452, 0.898027 56800 0.898027
0.90625]> 0.806452]>
Mr. Bhowmick 41 <41, 0.625> <41, [0.625,. 0.79057 13 <13,0.709677> <13,[0.709677, 0.842424 46500 0.79057
0.625]> 0.709677]>
Mr. Banerjee 39 <39, 0.5625 > <39, [0.5625, 0.75000 12 <12,0.677419> <12,[0.677419, 0.823055 42000 0.75000
0.5625]> 0.677419]>
Mr. Chattopadhyay 53 <53, 1> <53, [1,1]> 1 20 <20,0.935484> <20,[0.935484, 0.967204 80900 0.967204
0.935484]>
(Continued)
Table 25.7 Similarity Measures for Query 2 with Fuzzy Model. (Continued)
Vague Fuzzy SM of
Fuzzy represen represen SM of age represen Vague represen experience
Age tation of tation of with <53, Experience tation of tation of with Remuneration SM
Name (years) age age [1,1]> (years) experience experience <22,1,1]> (₹) (tuple)
Mr. Ganguly 52 <52, 0.96875> <52, [0.96875, 0.98425 19 <19,.0.903226> <19,[.0.903226, 0.950382 80200 0.950382
0.96875]> .0.903226>
Mr. Kher 56 <56, 0.90625 > <56, [0.90625, 0.95197 21 <21,0.967742> <21,[0.967742, 0.983739 81100 0.95197
0.90625]> 0.967742]>
Mr. Mukhopadhyay 54 <54, 0.96875 > <54, [0.96875, 0.9842 23 <23,0.967742> <23,[0.967742, 0.983739 82200 0.983739
0.96875]> 0.967742]>
Mr. Patnaik 55 <55, 0.9375> <55, [0.9375, 0.96825 27 <27,0.838710> <27,[0.838710, 0.915811 82000 0.915811
0.9375]> 0.838710]>
Mr. Sahoo 62 <62, 0.71875> <62, [0.71875, 0.84779 35 <35,0.580645> <35,[0.580645, 0.762001 122000 0.762001
0.71875]> 0.580645]>
Table 25.8 Resulting tuples for Query 2 with Fuzzy Model at α = 0.95.
Age Experience Remuneration
Name (years) (years) (₹)
Mr. Ganguly 52 19 80200
Mr. Kher 56 21 81100
Table 25.9 Resulting tuples for Query 2 with Fuzzy Model at α = 0.91.
Mr. Barik 51 18 58000
Mr. Bose 57 22 77900
Mr. Ghosh 49 17 57400
Mr. Ganguly 52 19 80200
Mr. Kher 56 21 81100
Mr. Patnaik 55 27 82000
It may be noted that as Query 2 involves two different attributes, the

intersection of the similarity measures of the two attributes give the sim-
ilarity measure of the corresponding tuple. The given query is examined
at different threshold values provided by the decision maker. The results
generated at the α-cut values 0.95 and 0.91 are demonstrated in Tables 25.8
and 25.9, respectively.
Next, we present the solution using a vague set for the same query.
Solution with Vague Model:

Once again, Algorithm 25.3.1 and Definition 25.2.3.1 are employed to
obtain the vague representation of the EMPLOYEE relation corresponding
to the Query 2, as presented in Table 25.10.
Table 25.10 Similarity measures for Query 2 with Vague Model.
SM of age Vague SM of experience

Age Vague representation with <53, Experience representation with <22, Remuneration SM
Name (years) of age [1,1]> (years) of experience [1,1]> (₹) (tuple)
Mr. Roy 30 <30, [0.28125, 0.537289 4 <4,[0.419355, 0.647989 32000 0.537289

0.440416]> 0.544234]>
Mr. Roychowdhury 31 <31, [0.3125, 0.485216]> 0.574358 6 <6,[0.483871, 0.695474 32500 0.574358
0.560575]>
Mr. Pramanik 34 <34, [0.40625, 0.640691 8 <8,[ 0.548387, 0.739961 34000 0.640691
0.516927]> 0.603663]>
Mr. Barik 51 <51, [0.9375, 0.953667]> 0.964182 18 <18,[0.870968, 0.931279 58000 0.931279
0.888225]]>
Mr. Bose 57 <57, [0.875, 0.878291]> 0.929415 22 <22,[1,1]> 1.000000 77900 0.929415
Mr. Ghosh 49 <49, [0.875, 0.875770]> 0.930180 17 <17,[0.83871, 0.910255 57400 0.910255
0.884063]>
Mr. Samanta 50 <50, [0.90625, 0.951472 16 <16,[0.80645, 0.888711 56800 0.888711

0.921369]> 0.839616]>
Mr. Bhowmick 41 <41, [0.625, 0.696428] > 0.785632 13 <13,[0.70967, 0.841028 46500 0.785632
0.751489]>
Mr. Banerjee 39 <39, [0.5625, 0.681963]> 0.743212 12 <12,[0.67741, 0.820811 42000 0.743212
0.704832]>
(Continued)
Table 25.10 Similarity measures for Query 2 with Vague Model. (Continued)
SM of age Vague SM of experience
Age Vague representation with <53, Experience representation with <22, Remuneration SM
Name (years) of age [1,1]> (years) of experience [1,1]> (₹) (tuple)
Mr. Chattopadhyay 53 <53, [1.000000, 1.000000 20 <20,[0.935484, 0.963133 80900 0.963133
1.00000]> 0.942255]>
Mr. Ganguly 52 <52, [0.96875, 0.982317 19 <19,[.0.90326, 0.945563 80200 0.945563
0.977372]> 0.914820>
Mr. Kher 56 <56, [0.90625, 0.946656 21 <21,[0.967742, 0.983458 81100 0.946656
0.914506]> 0.975348]>
Mr. 54 <54, [0.96875, 0.982701 23 <23,[0.96774, 0.982917 82200 0.982701
Mukhopadhyay 0.976538]> 0.971200]>
Mr. Patnaik 55 <55, [0.9375, 0.943755]> 0.966286 27 <27,[0.83871, 0.908944 82000 0.908944
0.873021]>
Mr. Sahoo 62 <62, [0.71875, 0.835870 35 <35,[0.58064, 0.756300 122000 0.756300
0.783282]> 0.667873]>
The resultant tuples generated after processing this imprecise query at

the same threshold or α-cut values are presented below in Tables 25.11
and 25.12.
Tables 25.8 and 25.11 confirm that when the threshold value is set to
0.95, the vague model produces finer results. The tuples of Mr. Ganguly
and Mr. Kher are not returned by the vague SQL, as it may be noticed
that their experience or age are not close enough to the said query. The
results shown in Tables 25.9 and 25.12 for the α-cut value 0.91 re-estab-
lish the fact that the vague SQL performs more effectively than its fuzzy
representation. It may be noted that the tuple of Mr. Patnaik, who is 55
years old and has 27 years of experience, was not retrieved utilizing the
vague framework.
Mr. Barik 51 18 58000
Mr. Bose 57 22 77900
Mr. Ghosh 49 17 57400
Mr. Ganguly 52 19 80200
Mr. Kher 56 21 81100
25.5 Conclusion
An algorithm has been designed in this research work to generate the fuzzy
or vague representation of attributes for an imprecise query. The suggested
technique produces a membership function for calculating membership
values that do not depend on the type of the attribute and are free of vari-
ous drawbacks identified in existing membership functions studied in the
literature. Using an Employee database, a comparison of fuzzy and vague
sets for handling imprecise queries has been performed. The results of
this study confirms that a database model based on vague sets can process
uncertain queries more efficiently than a fuzzy model.
References
1. Bosc, P. and Pivert, O. (1995). SQLF: A relational database language for fuzzy
querying. IEEE Transaction on Fuzzy Systems, Vol. 3, No. 1, pp. 1-17.
2. Chen, S. M. (1997). Similarity Measure between Vague Sets and between
Elements. IEEE Trans. Systems. Man and Cybernetics, Vol. 27, No. 1,
pp. 153-158.
3. Dutta, A. K., Idwan, S. and Biswas, R. (2009). A Study of Vague Search to
Answer Imprecise Query. International Journal of Computational Cognition,
Vol. 7, No. 4, pp. 63-69.
4. Gau, W. L. and Buehrer, D. J. (1993). Vague Sets. IEEE Trans. Syst. Man,
Cybernetics, Vol. 23, No. 2, pp. 610-614.
5. Hong, D. H. and Kim, C. (1999). A Note on Similarity Measures between
Vague Sets and between Elements. Information Sciences, Vol. 115, pp. 83-96.
6. Intan, R. and Mukaidono, M. (2000). Fuzzy functional dependency and its
application to approximate data querying. In Proceedings of International
Database Engineering and Applications Symposium, pp. 47-54.
7. Li, F. and Xu, Z. (2001). Measures of Similarity between Vague Sets. Journal
of Software, Vol. 12, No. 6, pp. 922-927.
8. Lu, A. and Ng, W. (2004). Managing Merged Data by Vague Functional
Dependencies, Lecture Notes in Computer Science, Vol. 3288, pp. 259-272.
9. Lu, A. and Ng, W. (2005). Vague Sets or Intuitionist Fuzzy Sets for Handling
Vague data: Which One Is Better? Lecture Notes in Computer Science, Vol.
3716, pp. 401-416.
10. Ma, Z. M. and Meng, X. (2008). A Knowledge-Based Approach for
Answering Fuzzy Queries over Relational Databases, Lecture Notes in
Artificial Intelligence, Vol. 5178, pp. 623-630.
11. Mama, R. and Machkour, M. (2021). Fuzzy querying with SQL: Fuzzy view-
based approach. Journal of Intelligent and Fuzzy Systems, Vol. 40, No. 5,
pp. 9937-9948.
12. Mishra, J. (2014). An Extension of Fuzzy Relational Database Model into
Vague Relational Database Model, Ph.D. Thesis, West Bengal University of
Technology.
13. Mishra, J. and Ghosh, S. (2014). Uncertain Query Processing using Vague
Set or Fuzzy Set: Which One Is Better? International Journal of Computers
Communications & Control, Vol. 9, No. 6, pp. 730-740.
14. Moreau, A., Pivert, O. and Smits, G. (2018). Fuzzy Query by example, 33rd
ACM/SIGAPP Symposium on Applied Computing (SAC-2018), France.
15. Nakajima, H., Sogoh, T. and Arao, M. (1993). Fuzzy Database Language and
Library- Fuzzy Extension to SQL. Proceedings of Second IEEE International
Conference on Fuzzy Systems, Vol. 1, pp. 477-482.
16. Raju, K.V.S.V.N. and Majumdar, A. K. (1988). Fuzzy functional dependen-
cies and lossless join decomposition of fuzzy relational database system,
ACM Transactions on Database Systems, Vol. 13, No. 2, pp. 129-166.
17. Srivastava, A., Yadav, S., Srivastava, N. and Khan, Z. (2016). Fuzzy Query: An
Impression in Query processing, Proceedings of IEEE Sponsored International
Conference on Advancement in Computer Engineering and Information
Technology (ACEIT 2016).
18. Takahashi, Y. (1993). Fuzzy database query languages and their rela-
tional completeness theorem, IEEE Transactions on Knowledge and Data
Engineering, Vol. 5, pp. 122-125.
19. Yadav, R. S. (2019). A Study of SQL query processing using soft comput-
ing techniques: a hybrid vague logic approach, International Journal of
Information Technology, Vol. 11, pp. 393-405.
20. Zadeh, L. A. (1965). Fuzzy Sets, Information and Control, Vol. 8, No. 3,
pp. 338-353.
21. Zhao, F.and Ma, Z. M. (2009). Vague Query Based on Vague Relational
Model, Advances in Computational Intelligence - Part of AINSC Book Series,
Vol. 116, pp. 229-238.
26
Estimating Error of Signals by Product
Means (N , pn , qn) (C, 2) of the Fourier
Series in a W(Lr, ξ (t))(r ≥ 1) Class
Pankaj Tiwari* and Aradhana Dutt Jauhari†
Division of Mathematics, Department of Basic Sciences, Galgotias University,

Greater Noida, G. B. Nagar, U.P., India
Abstract
The goal of this paper is to approach a new result on the degree of Fourier series
r
approximation of a function g ∈ W(L , ξ(t))(r ≥ 1) class by (N , pn , qn ) (C, 2) prod-
uct summability. Signals are handled as one-variable functions, while images are
represented as two-variable functions. The concept’s research is directly tied to the
rapidly developing field of information technology. The approximation theory is
a trigonometric polynomial approximation. Many academics have researched in
similar lines, but our research also proves some fresh findings.
Keywords: Degree of approximation, (N , pn , qn ) (C, 2)-means,

weighted generalized lipschitz class, Fourier series, Lebesgue integral
26.1 Introduction
Let L = L(0, 2π) denote the space of 2π- periodic and Lebesgue integrable
functions on the (0, 2π) Fourier series at point x:
∞
a
g 0 +
2 ∑(a + cosnx + b sinnx)
n =1
n n
Let ∑an be an nth partial sum sequence {sn} in an infinite series [9].
*Corresponding author: tiwaripankaj9716@gmail.com

†
Corresponding author: draradhana27@gmail.com
385
26.1.1 Definition 1
The transformation from sequence to sequence [5]:
N 1 n
t=
n ∑
= k 0 pn−k qk sk (26.1)
Rn
The (N , pn , qn ) mean of the sequence {sn} generalized by the series is

defined by the sequence {t nN }.
26.1.2 Definition 2
Riesz means can also be expressed as:
1 n
tnN == ∑ k =0 pkqk sk (26.2)
Rn
where Rn = p0 q0 + p1q1 + ⋯ + pnqn(≠ 0), p(−1) = q(−1) = R(−1) = 0.

The series ∑an is (N , pn , qn ) and is then summable to s.
We have the following regularity conditions for Riesz summability [10]:
pk qk
1. → 0, For each integer k ≥ 0, as n → ∞.
Rn
n
2. |∑ k =0 pk qk | < C |Rn |, where C can be any positive integer irre-
spective of n.
26.1.3 Definition 3
The (C,2) summability nth partial sum of the (C,2) transform is given [3]-.
2
Cn2 = ∑nk =0 (n − k + 1)sk → s as n → ∞ (26.3)
(n + 1)(n + 2)
By using the (C,2) method, the infinite series ∑an can be summable to
the definite as the number.
Estimating Error of Signals 387
26.1.4 Definition 4
(N, pn , q n ) of {sn} is a composite transformation over (C,2) as
∑
1 n
TnNC = pkqkCk2
Rn k =0
(26.4)
2 n pkqk (n − k + 1)
TnNC
= ∑ k =0 sk
Rn (k + 1)(k + 2)
If TnNC → s is n → ∞, then ∑an can be summable to s using (N , pn , qn )

(C,2) means.
26.1.5 Remark 1
If qn = 1, then in (26.2) the (N , pn , qn ) summability method reduces with
pn = 1(N , pn ) summability and it reduces to (N , qn ) summability.
Also, the degree to which of the function g: R → R is approximated by a
trigonometric polynomial tn of degree n:
||tn – g(x)||∞ = sup|tn(x) – g(x)| (26.5)
A function f(x) ∈ Lipα, if
|g(x + t) – g(x) |=O (|t|α) for 0 < α ≤ 1 (26.6)
and g(x) ∈ Lip(α, r), for 0 ≤ x ≤ 2π, if
1/r
 2π

∫ O(|t|α ), 0 < α ≤ 1, r ≥ 1, t > 0 (26.7)
r
 |( g (x + t ) − g (x ))
= | dx 
 0 
This is also applicable to any function that increases in a positive way

ξ(t), g(x) ∈ Lip(ξ(t), r), if
(∫ )
2π 1/r
0 )) |r dx
|( g (x + t ) − g (x= O(ξ (t )), r ≥ 1,t > 0 (26.8)
This is also applicable to any function that increases in a positive way

r
ξ(t), an integer r ≥ 1, f ∈ W(L , ξ(t))
(∫ )
2π 1/r
β
0 |( g (x + t ) − g (x ))}sin= x |r dx O(ξ (t )), β ≥ 0 ,t > 0. (26.9)
We will use the notations below as a guide:
ψ=
(t ) g (x + t ) + g (x − t ) − 2 g (x )
1 p q (n − k + 1) sin2 (n − k + 2)t /2
=
Pn (t ) ∑nk =0 k k sk (26.10)
2π Rn (k + 1)(k + 2) sin2t /2
r
Then, the weighted W(L , ξ(t)) class is a subset of the Lipα, Lip(α, r), and
Lip(ξ(t), r) classes. As a result, the following additions have been made:
Lipα ⊂ Lip(α, r) ⊂ Lip(ξ(t), r) ⊂ W(Lr, ξ(t))
for all 0< α ≤ 1 and r≥1.
26.2 Known Result

In 2011, Nigam [1] proved a theorem on (C,1)(E,q) means of the Fourier
series. Proceeding the work in 2014, Mishra et al. [4], the product (E,s)
(N,pn, qn) - summability mean of the Fourier series showed a theorem on the
degree of approximation. Pradhan [5] proposed the following theorem in
2016.
26.2.1 Theorem 1
If f ∈ W(Lr, ξ(t)) class is a 2π-periodic function that is integrable in the
Lebesgue sense in [0,2π], then the degree of approximation is given by
 β+
1
 1 
Tn
NE
O  (n + 1) r ξ 
− gr = 
  (n + 1)  
where TnNE is the (N , pn , qn ) (E,s) transform of {sn}, provided ξ(t) has the
following requirements: { }
ξ(t )
t
is a decreasing sequence,
1 r 1/r
 (n +1)
 t |φ (t )|    1 


∫ 0
 ξ (t ) 
 
sin βr
t dt  = O
 n + 1
r 1/r
 π
 t −δ |φ (t )| 


∫ 1
(n +1)
 ξ (t )  dt  = O(n + 1)

δ
26.3 Main Theorem

Various mathematicians including Nigam [2], Deger [6], and Mishra et al.
[7, 8] studied the degree of approximation using various summability
approaches. By using the Riesz-Cesaro product summability approach, we
r
obtain a novel result on the degree of approximation of function g ∈ W(L ,
ξ(t)) class.
The following theorem is established.
26.3.1 Theorem 2
1
If g ∈ W(Lr, ξ(t)) with 0 ≤ β ≤ 1 − , the degree of approximation by Riesz-
r
Cesaro product mean of the Fourier series satisfies for n = 0,1,2…
  1  β+ 
1
TnNC − g (x )r =
O ξ   n(n + 2) r
 (26.11)
  (n + 2)  
where TnNC is the (N , pn , qn ) (C,2) transform of sn, supposing that function

ξ(t) satisfies the required criteria,
{ }
ξ (t )
t
is a function that does not increase and
1 r 1/r
 (n + 2 )  t |ψ (t )sin β t|   n 

 ∫ 0

 ξ (t ) 
 
dt  = O 
 n + 2
(26.12)
r 1/r
 π
 t −δ |ψ (t )| sin β t  

 ∫ 1
(n + 2 )
 ξ (t ) 
 
dt  = O(n + 2)δ (26.13)
where δ is an arbitrarily defined s.t. qs(β − δ) – 1 > 0

r−1 + s−1 = 1, r ≥ 1 and (26.12) and (26.13) uniformly hold in x.
26.4 Some Auxiliary Results

Uniformly holding the following lemmas is necessary to prove the above
theorem.
26.4.1 Lemma-1
For
1 1 π π
0≤t ≤ , ≤ for 0 < t < ,
(n + 2) sint 2t 2
we get Pn(t) = O(n + 2)
n
pk qk (n − k + 1) sin2 (n − k + 2)t /2
Proof : |Pn (t )| =
1
2π Rn ∑
k =0
(k + 1)(k + 2)
sk
sin2t /2
1 n sk (n − k + 1) (n − k + 2)t 2 /π 2
≤ ∑ k =0
2π (k + 1)(k + 2) t 2 /π 2
= O (n+ 2)
26.4.2 Lemma-2
For (n + 2)−1 < t < π,
n
we get Pn (t ) = O( )
(n + 2)t 2
1 n pk qk (n − k + 1) sin 2 (n − k + 2)t /2
Proof : |Pn (t )| = ∑ k =0 sk
2π Rn (k + 1)(k + 2) sin 2t /2
t t
Using Jordan’s Lemma sin ≥ and sinkt ≤, we have
2 π
1  π 2  n pk qk (n − k + 1)
≤   ∑ k =0 sk
2π Rn  t 2  (k + 1)(k + 2)
n 
= O 
 (n + 2)t 2 
26.5 Theorem’s Proof

Using Riemann-Lebesgue theorem, we have
π
1 sin 2 (n − k + 2)t /2
sn ( g ) − g ( x ) =
2π ∫ 0
ψ (t )
sin 2t /2
dt
The Riesz Cesaro transform of the sequence is given by
1 p q (n − k + 1) π sin2 (n − k + 2)t /2
{TnNC − g (x )} =∑nk =0 k k sk ∫ 0 ψ (t ) dt
2π Rn (k + 1)(k + 2) sin2t /2
π
=
∫ ψ (t )P (t )dt
0
1
n
(n+2 ) π
= ∫0 ψ (t )Pn (t )dt + ∫ 1 ψ (t )Pn (t )dt
(n+2 )
= I1 + I 2 (26.14)

1
( n+ 2)
where |I1| ≤ ∫ 0 |ψ (t )||Pn (t )| dt
Applying Minikowski’s inequality, since
|ψ(x, t) – ψ(x)| ≤ |g(u + x + t) – g(u + x)| + |g(u – x – t) – g(u – x)|
2π 1/r
 
 ∫ 0
|ψ ( x + t ) − ψ ( x )sin x| dx 

β r
2π 1/r
 
≤
 ∫ 0
|g (u + x + t ) − g (u + x )sin x| dx 

β r
1/r
 2π 
+
 0∫ |g (u − x − t ) − g (u − x )sin β x|r dx 

= O{ξ (t )}
and the fact that

g ∈W ( Lr , ξ (t ) ) ⇒ ψ (t ) ∈W ( Lr , ξ (t ) ) and Lemma 1, we have
{ }
I /r 1/s
 1
  t |ψ (t )| sin β t r    1
 ξ (t )| Pn (t ) s  
∫ ∫
(n + 2) (n + 2)
I1 ≤    dt     β dt  
 0   ξ (t )     0  tsin t  
{ }
1 1/ s
n    ξ (t )|Pn (t )| s  
= O 
∫
(n + 2)
  dt   {using (26.12)}
 (n + 2)   0  tsin t
β
 
{ }
1 1/ s
n    
s
 ξ (t )O(n + 2)
= 
∫
(n + 2)
  dt   {using Lemma 1}
 (n + 2)   0  tsin β t  
Using the second mean value theorem for integrals, ξ(t) is a decreasing
function and we get
1 1/s
 1     dt  
 1
≤ O  nξ 
∫
(n + 2)
  0 ≤ε <
  (n + 2)    0  t ( β +1)s   (n + 2)
1 1/s
  1    
∫ [t −( β +1)s dt ]
(n + 2)
= O  nξ  
  (n + 2)    0 
1/(n + 2) 1/s
  1     [t 
− ( β +1)s +1
]
= O  nξ    
  (n + 2)     −(1 + β )s +1 ε 
− β −1/ r
 1   1 
= O  nξ    
  (n + 2)    (n + 2) 
 1  
= O ξ   n(n + 2)β +1/r  (26.15)
  (n + 2)  
Now,
π
|I 2 | ≤ ∫ 1 |ψ (t )||Pn (t )|dt
(n + 2)
2t
Using Hölder’s inequality as a starting point, ≤ |sint| < 1 [10]
π
1/r 1/s
  t −δ |ψ (t )|sin β t   
r
 ξ (t )|Pn (t )|  
s
π π
≤
 ∫ 
1/(n+2)  ξ (t )
 dt  ⋅ 
   ∫  −δ β  dt 
1/(n+2)  t sin t  
1/s
  n   
s
 π  ξ (t )O(  
δ   (n + 2)   
= O(n + 2) 
  ∫  t −δ + β +2  dt   using (26.4)
1/(n+2)  
{ }
1/s
  π
ξ (t )
s
 
= O(n + 2)δ −1n  
  ∫1/(n+2) t −δ + β + 2
dt   by Lemma (2)
 
we put t =1/x, dt = -(1/x2)dx
1/ s
   1 
s

 (n + 2)  ξ  
δ −1    x   dx  
= O{n(n + 2) }  
 ∫ 1/π
 x δ − β − 2  x 2  
1
ξ 
x
Because ξ(t) > 0 and are both rising functions, the second mean
(1 / x )
value theorem is used and
1/ s
 1    (n + 2)
dx 
= O n(n + 2)δ −1 ξ 


 (n + 2)     ∫1/π x δ s − β s −2s+2  

1/ s
 δ −1  1     x β s + 2 s −δ s − 2+1  
= O n(n + 2) ξ   
  (n + 2)     β s + 2s − δ s − 1  
(n + 2)
 1     x ( β −δ + 2)−1 s  
= O n(n + 2)δ −1 ξ 
  (n + 2)     ( β − δ + 2) − 1  
  
s   1
π
 1  
= O ξ   n(n + 2)β +1/r  (26.16)
  (n + 2)  
from (26.14) and (26.15), (26.16) becomes
  1  
TnNC − g (x ) r =
O ξ   n(n + 2)β +1/ r 
  (n + 2)  
26.6 Applications
From our main theorem, we can derive the following corollaries.
26.6.1 Cor. 1
If ξ(t) = tα, then weighted class W ( Lr , ξ (t ) ) , 1 ≤ r < ∞ reduces to the group
of class Lip(α, r) and then a function f is a degree of approximation belong-
ing to the class Lip(α, r), r−1 ≤ α ≤ 1, given by
TnNC − g (x )r =+
O{n(n 2)1/ r }
Proof: Setting produces, the result is β = 0 in (26.5).
26.6.2 Cor. 2
For r → ∞ in Cor. 1 for the class function, g ∈ Lip(α, r) reduces to the class
Lip α and a function g belonging to the class Lipα, 0 < α < 1 whose degree
of the approximation is given by
O{n(n 2)α }
TnNC − g (x )r =+
26.7 Conclusion
The previous finding on function degree of approximation demonstrated
that the result is frequently true in nature and can be simplified to a few
special cases. As a result, the current research can be used for a series of
issues in the fields of analysis, technology, and engineering. Figure 26.1
shows some unique behaviour in the Fourier series approximations [11].
1 n=1
0 t
−1
1
n=5
0 t
−1
1 n=11
0 t
−1
1 n=49
0 t
−1
Figure 26.1 Approximations of such Fourier Series to sq (t). Each plot indicates the
number of terms in the Fourier sum and the square wave is displayed as a dashed line over
two periods.
Acknowledgement
The sincere gratitude goes to the anonymous referees for their careful
reading, remarks, and valuable comments, as well as many other useful
suggestions for improved presentation. The author wishes to express his
appreciation to the members of the editorial board.
References
1. H. K. Nigam: “Degree of approximation of a function belonging to weighted
(Lip (ξ(t), r) class by (C,1) (E,q) means”, Tamkang Journal of Mathematics, 42,
31-37 (2011).
2. H. K. Nigam: “On Approximation of functions by product operators”, Survey
in Mathematics and its Applications ISSN 1842-6298(Electronic), 1843-
7265(print) Volume 8 (2013), 125-136.
3. K. Sharma. and S.S. Malik: “Degree of Approximation by Product means
of the Fourier Series in a (W ( Lr , ξ(t ) ) (r > 1) − class by (E,1)(C,2) product
summability Transform”, International Journal of Mathematical Trends
and Technology (IJMTT), Volume-8 Number-2,ISSN 2231-5373,2014,
10.14445/22315373/IJMTT-V8P518.
4. M. Mishra et al.: “Approximation of Fourier series of a function of Lipschitz

class by product means” Journal of advances in Mathematics, 9 2475-2484
(2014).
5. T. Pradhan et al.: “Approximation of signals belonging to generalised Lipschitz
class using ( N , pn , qn )( E , s ) -summability mean of Fourier series”, Cogent
Mathematics (2016),3: 1250343.http://dx.doi.org/10.1080/23311835.2016.1
250343.
 “On Approximation to functions in the W ( Lp , ξ(t ) ) class by a new
6. U. Deger:
matrix mean”, Novi Sad J. Math, Vol.46 No.1, 2016, 1-14.
7. V. N. Mishra et al.: “Approximation of Signals by Product Summability
Transform”, Asian Journal of Mathematics and Statistics 6(1): 12-22, 2013,
ISSN 1994-5418/ DOI: 10.3923/AJMS.2013,12.22.
8. V. N. Mishra, K. Khatri, L. N. Mishra: “Product Summability Transform
of Conjugate series of Fourier series”, Hindawi Publishing Corporation,
International Journal of Mathematics and Mathematical Sciences, Volume
2012, Article 1D 298923, 13 pg doi: 10.1155/2012/298923.
9. A. Zygmund: “Trigonometric Series”, second ed., Cambridge University
Press, Cambridge, 1959.
10. G. H. Hardy: Divergent Series Oxford University Press 1st Ed. 1949.
11. https://eng.libretexts.org/Bookshelves/Electrical_Engineering.
About the Editors
Biswadip Basu Mallik is presently a Senior Assistant Professor of

Mathematics in the Department of Basic Sciences & Humanities at the
Institute of Engineering & Management, Kolkata, India. He has been
involved in teaching and research for more than 21 years and has published
several research papers in various scientific journals and book chapters
with reputed publishers. He has authored five books at undergraduate
levels in the areas of Engineering Mathematics, Quantitative Methods,
and Computational Intelligence. He has also published five Indian
patents along with nine edited books. His fields of research work are
Computational Fluid Dynamics & Mathematical Modelling. Prof. Basu
Mallik is a Managing editor of the Journal of Mathematical Sciences &
Computational Mathematics (JMSCM), USA. He is also an Editorial
board member and reviewer of several international journals. He is a
senior life member of the Operational Research Society of India (ORSI)
and a life member of the Calcutta Mathematical Society (CMS), Indian
Statistical Institute (ISI), Indian Science Congress Association (ISCA), and
International Association of Engineers (IAENG).
Niranjanamurthy M., Ph.D., is working as an Assistant Professor at the

Department of Artificial Intelligence and Machine Learning, BMS Institute
of Technology and Management, Bangalore, INDIA. He completed
Ph.D. in Computer Science, M.Tech. - Master of Computer Technology
(Software Engineering), M.Phil.-Computer Science, MCA at VTU,
Belgaum, Karnataka, BCA from Kuvempu University with University 5th
Rank. Dr. Niranjanamurthy is having 13 years of teaching experience and
2 years of industry experience as a Software Engineer. He has published
20 books and 85 articles in various National/International Conferences/
International Journals. He has filed 30 Patents out of which 6 has been
granted. Four research scholars are pursuing Ph.D. work under his super-
vision in the area of Data Science, ML, and Networking. He is the reviewer/
editorial board member of 22 International Journals and is also a National/
International Ph.D. thesis examiner. Dr. Niranjanamurthy received the
397
398 About the Editors
Best Research journal reviewer award twice. He had also received ear-
lier Young Researcher award in Computer Science and Engineering. He
obtained Patent Award - 2021 from KSCST, Karnataka, India. He con-
ducted various National and International Conferences, National Level
workshops and delivered lectures. Dr. Niranjanamurthy is associated
with various professional bodies as follows: IEEE Member, Life Member
of “The International Association of Engineers” (IAENG), and Member
of “Computer Science Teachers Association” (CSTA). His areas of interest
are Data Science, ML, E-Commerce, and M-Commerce related to Industry
Internal tool enhancement, Software Testing, Software Engineering, Web
Services, Web-Technologies, Cloud Computing, Big Data Analytics, and
Networking.
Sharmistha Ghosh, Ph.D., is currently working as a Professor at the

Institute of Engineering & Management, Kolkata, India. She obtained
B.Sc.(H) and M.Sc. degrees in Mathematics from Jadavpur University,
Kolkata, India in the years 1992 and 1994 respectively as University Gold
Medalist. Dr. Ghosh received Ph.D. degree in Mathematics from Indian
Institute of Technology, Kharagpur, India in 1999. She also obtained M.Sc.
nat. degree in Industrial Mathematics from University of Kaiserslautern,
Germany in 2000 and is a DAAD fellow. Her major field of study includes
Fuzzy & Vague Databases as well as Computational Fluid Dynamics. She has
published several research articles in International Journals/Conferences,
book chapters in these domains and also one edited book. She is the editor
of renowned international journals and also works as reviewer of journals
as well as doctoral theses in India and abroad. She has delivered invited
talks in several national as well as international forum and organized many
national and international conferences and workshops.
Krishanu Deyasi, Ph.D., is currently an Associate Professor in the

Department of Basic Science & Humanities, Institute of Engineering
& Management Kolkata, India. He did Ph.D. from Indian Institute of
Science Education and Research (IISER) Kolkata. Dr. Deyasi obtained
his Master Degree in Mathematics from Indian Institute of Technology
(IIT), Kharagpur. He has Post-Doctoral experience from The Institute
of Mathematical Sciences (IMSc) Chennai. His major area of research is
Spectral Graph Theory and Complex Networks. He has written 3 books in
the area of Engineering Mathematics and has published one edited book.
He has published papers in National and International Journals. He is also
the editor of some Scientific Journals.
About the Editors 399
Santanu Das is currently working as an Assistant Professor at the

Department of Basic Science and Humanities at Institute of Engineering
& Management, Kolkata, India. He completed his under-graduation and
post-graduation in Mathematics from Jadavpur University and is cur-
rently pursuing his PhD from the same University. His field of research
work is Cosmology and Dynamical system. He is the Faculty coordinator
of Society for Data science IEM student chapter. He takes classes of various
Mathematics Courses for B.Tech,, M.Tech. and MCA students. He is also
a Development Content Editor of the Journal of Mathematical Sciences &
Computational Mathematics (JMSCM), USA.
Index
Accreditation, 79 Automorphism, 291, 293, 294, 296,

Accrues, 206 297, 300, 301, 303
Activation, 94 Average arc problem, 327, 328
Adaptive, 77, 79 Awareness, 78, 85, 101, 105, 111, 191
Adjoint equation, 41, 42 Ayurveda, 80, 81, 83, 84, 87
Adoption, 85 Ayushgrid, 79
AED, 193, 194, 198, 199, 202 Azadikaamrutmahotsav, 81
Agricultural, 49, 55
Ambient temperature, 224 Backward Propagation Model, 66
Amino Acid, 91, 92, 94, 96, 98 Balking, 151, 152, 153, 154, 163, 164,
Amino Acids Sequence, 91 165
Amplitude ratio, 127, 129, 130, 133 Bernoulli Schedule, 151, 153, 154, 163,
Analytical hierarchy process, 188 169
Animations, 140 Best fit distribution, 19, 20, 22, 24, 31
Ant Colony Optimization (ACO), 60 Bidirectional Long Short-Term
Anti-parallel, 239, 241, 242, 245, 246 Memory (BLSTM), 62
Approximation, 24, 153, 226, 239, 241, Big data, 3, 188
244, 245, 248, 385, 388, 389, 394, Binary Coding Scheme, 62
395 Biomarkers, 34
Arc problem, 327, 328, 330 Bio-Portal, 186, 192, 201
Arrival time, 155, 164, 165, 166, 169, Biot parameter, 116, 117, 131, 133
332 Blow up, 305, 306, 307, 317
Artificial Intelligence (AI), 1, 59, 60, Boundary condition, 38, 39, 115, 116,
63, 205, 266 124, 227, 228, 256, 257, 259, 315
Artificial Neural Networks (ANN), 60 Bounding box, 64
Asvagandha, 87 Breakdown Machines, 337, 347
Asymmetric, 173, 183, 239, 241, 244, Bright Cities, 139
245, 246, 247, 248, 250 Brownian motion coefficient, 224
Atmospheric boundary layer (ABL), Brownian motion parameter, 225, 228,
19, 24 230, 231
Attribute, 7, 50, 51, 53, 54, 55, 65, 188,
350, 364, 366, 367, 368, 369, 370, Capacity violation, 333
375, 378, 382 Catastrophic, 77
Aubin-Lions, 41 Cell count, 34, 43
401
402 Index
Cellular automata, 34 Conservativeness, 39

Central difference, 42 Conspicuous, 80
Centre, 79 Constant deterioration rate, 254
Certification, 83, 86, 140 Continuous-time, 242, 327, 328, 329,
Chang-Cooper, 42 331
Characteristics, 61, 62, 64, 78, 187, Contraflow, 239, 241, 242, 244, 247,
188, 189, 190, 192, 196, 213, 266, 248
270, 287, 342, 345, 375 Convective boundary layer height
Chemical reaction, 226, 231, 234, 235 (CBLH), 19, 20, 21, 23
Chemotherapy, 33, 34, 35, 45 Convergent, 40, 41, 42
Chi-square statistics, 22, 24, 31 Convex combination, 349, 350, 351,
Citizens, 82, 87, 139, 207, 208 358
Class Mapping, 190, 196, 198, 199, Conveyor, 210, 211, 212, 213, 214, 216,
200, 201 217, 218
Classification, 1, 2, 3, 4, 7, 11, 12, 13, Convolutional Neural Network (CNN),
14, 24, 49, 50, 55, 61, 62, 63, 65, 60, 61, 62, 65, 93, 209, 289
67, 68, 69, 70, 73, 92, 196, 209, Coronavirus, 101, 102, 218
210, 291 Correlation, 3, 65, 96, 144, 146, 147,
Client, 14, 151, 152, 153, 154, 155, 157, 187
163, 164, 165, 166, 169 Course Material, 140
Cluster, 7, 194, 196, 197, 198, 199, 200 Covanta, 207
Coercive, 40 COVID-19, 77, 78, 79, 80, 81, 82, 84,
Cohn-Kanade expression, 68 87, 88, 101, 102, 205, 209, 254
Coil, 92 Criteria, 8, 67, 187, 188, 189, 190, 191,
Collaboration, 84, 85, 87, 141, 206, 207 192, 194, 197, 198, 270, 277, 350
Colon cancer, 33, 34, 35, 44 Cropping, 61, 63, 64, 67, 284, 286
Combination therapies, 33, 34, 35, 44 Crowd motion, 35, 42
Comparative analysis, 70, 74, 363, 368 Crust, 116
Complex Network, 101, 106 Cullpdb, 91, 92, 93, 96, 97
Complexity, 78, 240, 246, 266, 285, 329 Cytotoxic effects, 34
Composite, 116, 387
Computational analysis, 101, 111 Dashboard, 80, 83
Computer networks, 60 Data Analysis, 13
Computer vision, 284 Data Analytics, 3
Computers, 138, 139, 206 Data Visualization, 8
Concentration boundary layer, 227 Data1199, 91, 92, 93, 97
Concentration gradient, 234 Database Model, 363, 364, 368, 382
Confused Memory, 192, 193, 194, 199, Data-driven, 8, 188
202 DBT, 83
Congestion, 151, 153, 154, 155, 164, Decision level hybrid method, 62
169, 329 Decision Tree, 49, 50, 51, 53, 54, 55
Congestion dependent, 151, 153, 154, Deep learning, 59, 60, 62, 94, 96
155, 164, 169 Degree of Approximation, 388, 389,
Conservation constraint, 332 394
Index 403
Density of the fluid, 224 Elastic parameter, 117, 224, 231, 235,
Despite, 85, 97, 139, 206, 207 236
Detection, 1, 11, 12, 13, 14, 20, 22, 30, E-Learning, 137, 138, 139, 140, 141,
33, 49, 51, 54, 59, 60, 61, 62, 63, 142
64, 67, 77 Electrical conductivity, 224
Differential equations, 103, 228, 256, Electronic, 80, 139, 141, 209, 251, 252
257, 259 Encoder-Decoder, 5, 91, 92, 94, 95, 96
Digital, 2, 55, 82, 83, 85, 88, 139, 141, Energy ratios, 126, 131, 134
142, 209, 248 Entropy, 49, 50, 51, 52, 53, 54, 96, 224,
Digital Transformation, 139 226, 286, 287
Dimensionless elastic parameter, 224 Environment, 11, 21, 28, 62, 79, 140,
Dimensions, 240, 267, 273 142, 152, 173, 175, 183, 184, 205,
Directives, 80 206, 207, 215, 218, 273, 349
Disaster, 80 Epilepsy disease, 187, 191, 192, 201,
Disease, 6, 34, 49, 50, 51, 52, 53, 54, 55, 203
85, 101, 102, 111, 187, 189, 191, EPILONT, 191, 192, 193, 194, 198, 199,
192, 198, 201, 202, 203, 207, 284 200, 201, 202
Distributed, 25, 81, 151, 154, 155, 166, EPSO, 191, 192, 193, 196, 198, 199,
169, 176, 208, 337 200, 201, 202
Diverse, 59, 84, 85, 154, 239, 284 Equipment, 141, 206
Doctors, 192, 284, 285 ESSO, 193, 194, 196, 198, 199, 200,
Domain knowledge, 188 201, 202
Dosages, 34, 36, 43, 44, 45 Evaluation, 1, 11, 52, 69, 116, 140, 147,
Douglas-Gunn, 42 187, 188, 189, 192, 203, 276, 286,
Doxorubicin, 33, 35, 36, 43, 44 343
Dynamic, 22, 28, 33, 34, 35, 36, 37, 41, EvoMSA, 62
45, 101, 102, 103, 106, 107, 111, Existence, 39, 116, 139, 218, 223, 305,
173, 174, 175, 176, 178, 183, 209, 306, 307, 314, 333, 349
226, 240, 241, 242, 327, 328, 329, Exploitation vector, 68
330, 331, 333, 344
Dynamic Model, 28, 34, 35 Facebook, 14, 85, 86, 138, 139
Dynamic network, 327, 329, 330 Face-To-Face Communication, 142
Dynamic viscosity, 224 Failed Machines, 337, 338, 340, 347
Dynamical systems, 34 Feasibility of flow, 331, 332, 333, 334
Fibonacci Search Method, 265, 268,
E-Business, 140 270, 271
E-Citizens, 139 Finite difference, 42
Education, 1, 14, 82, 83, 84, 139, 140, Flickr, 138, 139
141 Flow capacity, 327
EduSERE, 62 Fluid, 116, 117, 223, 224, 225, 226
Efficacy, 80, 84, 218, 364 F-measure, 59, 69, 70, 71, 72, 74
E-Government, 140 Focal, 193, 194, 196, 198, 202
E-Health, 140 Fokker-Planck, 33, 35, 38, 42
404 Index
Fourier series, 385, 388, 389, 394, 395 ILE-JAVA, 62

Free surface, 115, 116, 121 Image pre-processing 63
Friedman Rank Test, 268, 273, 276 Image segmentation 267, 283, 289
Fuzzy Logic, 51 Immune pathways 34
Fuzzy set, 350, 351, 363, 364, 365, 366 Immunity 80, 81, 82, 87, 102
Immunotherapy 33, 35, 45
Garbage, 206, 207, 208, 209, 210, 213, Impatience 151, 152, 153, 169
215, 216, 218 Impedance 115, 116, 124, 129, 130,
Genetic algorithm, 266, 337, 338, 344, 131, 133
347 Imprecise Query 368, 381, 382
Genetic toggle switch, 173, 180 Incident 115, 116, 121, 122, 127, 128,
Genomics 84 129, 131, 132, 133
Geographical Location, 142 Incinerator 205, 210, 211, 212, 213
Geriatric, 81 Information Gain 49, 50, 51, 54, 55
Gestures, 60, 63 Infusion 87
Ghee, 87 Inhalation 87
Gray scaling, 61 Initial condition 36, 180
Growth Stage, 51, 52, 53, 55 Initiatives 77, 78, 82
Guduchi, 87 Inlet 210, 211, 212, 213
Innovation 78, 141, 212
He@lthy, 85 Insulated 116, 125
Head Trauma, 193, 194, 199, 202 Integration Of Communication
Health, 13, 33, 50, 77, 79, 80, 81, 82, Networks 139
83, 85, 86, 87, 101, 102, 205 Intelligently Evaluating 138
Health Care, 205 Intelligibility 60
Healthcare, 13, 60, 78, 79, 81, 82, 205, Interdisciplinary 83, 84
206, 218, 219 International 79, 84, 86, 191
Heap coupled bat Optimization Internet 2, 138, 139, 140, 141, 142, 152,
(HBO), 59 206, 209, 210
Helix, 92, 269 Internet Of Things 139, 140, 141, 142,
Hierarchy, 188, 192, 194, 210 206, 209, 210
High-Bandwidth, 140 Intervention 77, 79, 80, 84, 88, 102
Homotopy analysis, 225 IoT 137, 138, 139, 140, 141, 142, 144,
Hybrid 59, 61, 62, 65, 74, 187, 202, 145, 147, 205, 207, 208
267, 333, 334 Isothermal 115, 116, 125, 127, 224
Hybrid Deep Convolutional based Isotropic 115, 117, 121, 122, 126, 133
Golden Eagle Network (HDC- Iterative Dichotomiser 3 49, 50
GEN), 59
Hybrid network, 65, 333, 334 Jupyter 11
Hyper-Connection Linking, 139
Hypothesis testing, 22 Keras 96
Kernel function 66, 307
IEEE CEC’2019, 276 Kinematic coefficient of viscosity 224
IL-2, 33, 35, 36, 43, 44 K-Nearest neighbors (KNN) 60
Index 405
Knowledge management 13 Midnight, 81

Kolmogorov-Smirnov (KS) test, 31 Minimization problem, 39, 41
Kwath, 87 Minimum time, 331
Mobile Communication, 210
Lane reversal, 241, 243, 244, 245, 246, Model, 2, 3, 4, 5, 8, 9, 10, 11, 12, 34,
247, 248 35, 42, 61, 62, 63, 66, 68, 70-74,
Lane, 239, 241, 243, 244, 245, 246, 247, 92-97, 102-110, 148, 152, 154,
248 155, 163, 164, 178, 179, 189,
Leadership, 77, 79 209, 242, 252-261, 285-288, 308,
Leaf Blast, 49, 51, 55 329, 331, 338, 343, 364, 368,
Learning Tools, 140 369-381,
Length-bounded, 140, 245, 248 Moderators, 338, 339
Lesions Type, 54 Molecular diffusivity, 224
Lewis number, 225, 228 Momentum boundary layer, 227
Lidar Backscatter signals, 19 MSM, 206
Linear programming, 327, 331 Multi-commodity, 241, 242, 244, 245,
Linguistic Variable, 51, 52 246, 247
LinkedIn, 141 Multi-commodity flow, 241, 245, 246
Lipschitz domain, 38 Multi-criteria decision
Logarithmic Sobolev inequality, 313, making (MCDM), 188
317, 321, Multiserver system, 337
Long Short-Term Memory Network,
96, 97 NAMASTE, 85
Longitudinal, 116, 121, 122, 127, 128, Nanofluids, 223, 225
129, 130, 131, 132, 133 Natural Killer cells, 34
Lymphocytes, 35, 36 Natural Language Processing, 2, 5, 7,
14, 92
Machine Interactions, 138, 141 Naturopathy, 84, 87
Machine learning, 4, 5, 7, 8, 9, 11, 13, Nave Bayes, 60
51, 62, 208, 209, 266, 284 Nehari functional, 311
Magnetic parameter, 224, 230, 231, Network, 5, 14, 60, 61, 62, 65, 68, 85,
235, 236 92, 93, 94, 96, 97, 106, 111, 139,
Mannkibaat, 87 140, 152, 208, 209, 210, 214, 240-
Materials, 116, 140, 141, 219 248, 266, 284, 285, 289, 328, 329,
Mathematical Model, 34, 102, 225, 330, 333, 334,
242, 248, Neural cell, 283
MATLAB, 63, 64, 68, 127, 273, 345 Neural Network, 60, 61, 92, 93, 209,
Maximum dynamic, 241 284, 289
Maximum flow, 240, 241, 329 Neurological disorder, 190, 284
MCU, 210, 212 Neurologists, 201
Medical Science, 83, 84, 225 Nickelodeon, 86
Meditation, 83 Non-abelian groups, 291, 292, 293
Metrics, 68, 69, 70, 74, 96, 154, 163, Non-conservation constraint, 332
164, 188, 189, 201, 268, 286 Non-dimensional temperature, 224
406 Index
Non-symmetric, 241, 244, 245, 246, Pores, 116

247, 248 Porosity parameter, 224, 234
NP-complete, 244, 245, 333, 334 Porosity, 116, 117, 133, 234
Nusselt number, 230 Porous, 116, 117, 121, 122, 126, 127,
133, 224, 227, 234, 236
One Hot Encoding, 95 Porous medium, 126, 227
Ontology Engineering, 189 Potential well, 309
Ontology Selection, 188, 189, 190, 202 Potential, 35, 77, 118, 119, 141, 142,
Open-loop, 40 308, 309, 364
Optimal control, 35, 45 Pranayama, 83
Optimality condition, 41 Prandtl number, 224, 228, 231, 235
Optimality system, 41 Preprocessing, 2, 3, 4, 9, 14, 61, 62, 64,
Optimistic, 69 67, 94
Optimization, 35, 39, 42, 60, 65, 96, Primary Structure, 92
208, 252, 266, 267, 268, 270, 273, Probabilities, 68, 69, 153, 155, 157,
277, 279, 344 162, 163, 169
Probability density function, 35, 154
Groups, 291, 292, 293 Prophylactic, 83
Proteins, 91, 92, 176, 177, 178
Pandemic, 78-86, 102, 103, 206, 207, Protocol, 83, 85, 87
215, 218 Psychological, 60
Paradigm, 80, 154 Psychosocial, 87
Parallel computation, 60 Python, 96, 106
Parameter estimation, 35
Particle Swarm Optimization (PSO), 60 Q3 Accuracy, 93, 96, 97
PARTITION, 245, 332, 333, 334 Query processing, 364, 367
Peptide, 92 Questionnaire, 142
Performance, 11, 61, 63, 66, 68, 69, 70, Queue, 152, 153, 154, 163, 166, 169,
74, 92, 96, 97, 154, 163, 164, 208, 337, 338, 343, 347
267, 268, 273, 277, 285, 286, 287, Quickest, 240, 241, 244, 245, 246, 247,
342, 358, 366 248, 329
Personalized therapies, 34
Pharmaceutical, 84, 102, 207 Radiation parameter, 224
Pharmacokinetic models, 34 Random Forest Algorithm, 55
Pipeline, 102 Randomness, 35, 45
Pitchfork bifurcation, 174-183 Reflecting barrier, 38
p-Laplacian type, 306 Reflection, 116, 121, 122, 127, 128,
Plate concentration parameter, 224, 130, 131, 133
234, 236 Related Works, 207
Plate temperature parameter, 224 Relaxation time, 115
Polynomial, 240-245, 247, 248, 328, 329 Reneging, 152, 154, 163, 164, 166
Polynomial-time, 240, 241, 242, 245, Repairmen, 338
247, 248 Resource allocation, 177, 178, 180, 182
Populations, 35, 36, 103, 104, 107, 273, Restorative, 218
Index 407
Rice Crop, 49, 51, Strength of uniform magnetic field,

Rocks, 116 224
Stress, 80, 83, 116, 117, 277
Saddle node bifurcation, 174, 176, 177, Stretching surface, 225
180, 182, 183 Strongly polynomial, 241, 248, 329
Scalability, 139 Susceptive, 52, 54, 55
Screen, 210 Symmetric, 179, 180, 241, 244, 246-248
Segmentation, 51, 284-289 Symptoms, 33, 49, 50, 51, 52, 55, 84,
Seismology, 116 191, 192, 194, 195, 200, 202
Seizure, 191, 194, 198, 200, 201 System Architecture, 210, 211
Semantic Web, 138, 139, 188
Sensors, 140, 141, 207-214, 216, Technicians, 339
SentiTEXT, 62 Temperature gradient, 231
Sequence-Labeling, 92 Temperature sensor, 138
Sherwood number, 230, 231 Temporal variation of lidar signals,
Similarity Measure, 364, 365, 366, 370, 24, 27
371, 373, 374, 375, 376, 377, 378, Tensor flow, 96
379, 380 Thermal boundary layer, 227
Single valued neutrosophic fuzzy Thermal conductivity of the fluid, 224
numbers, 351 Thermodynamics, 224
Sink, 235, 241, 244, 328, 329, 332, 333 Thermoelastic, 115, 116, 121, 122, 126,
Skin-friction coefficient, 230 133,
Smart Bins, 217 Thermophoresis parameter, 225, 228,
Softmax, 67, 68, 95 230, 231
Soil, 51, 116, 208, 218 Thermophoretic diffusion coefficient,
Solid, 115-117, 133, 206, 207, 209, 210, 224
212 Threshold, 21, 138, 153, 209, 347, 373,
Soret number, 230 378, 381
Source, 3, 6, 83, 86, 96, 142, 143, 145, Time dependent holding cost, 254
147, 191, 192, 206, 225, 241, 244, Time expansion, 241, 245
306, 328, 329, 330, 332 Time horizon, 253, 255, 329, 333
Specific heat, 224 Time-expanded network, 328, 330,
Split Rollers, 210 333, 334
SQL, 226, 228, 364, 367, 373, 375, 381 Toxicity, 34, 44, 45
SSA-DCNN, 70-73 Transit time, 241, 244, 246, 248, 328,
State equation, 35, 41, 340 329, 330, 333, 334
Static, 96, 138, 187, 240, 242, 245, 329 Transverse, 121, 127, 129, 130, 131,
Stationary Distribution, 154 133, 231, 235, 266
Statistical Techniques, 21 Trapezoidal demand, 252, 254
Stericycle, 207 Trash, 207-210, 212, 215, 216, 218
Stochastic, 35, 37, 38, 42, 44, 45, 268 TSD problem, 277
Storage capacities, 329 Tumorigenesis, 34
Strand, 92 Turbulence, 19, 20, 25, 26, 28
Stream function, 224, 227 Twitter, 14, 85, 86, 141
408 Index
Unani, 78, 80, 83 Web 2.0, 138, 139

User Interface, 8 Web 3.0, 138, 139
Weight vector, 68
Vacation Interruption, 153, 154, 166, 169 Weights, 3, 96, 188, 189, 190, 193, 197,
Vacations, 152, 153, 163 198, 201, 202, 285, 286
Vaccination, 102-111 Width window, 66
Vague set, 364-367, 378, 382 Wiener process, 38
Vague SQL, 381 Working Vacations, 152, 153,
Velocity, 224, 226, 231, 234, 235, 236 World wide web, 138
Video Conferencing, 139, 140 Wrapping, 212
Virtual Learning, 141
Viscoelastic, 223, 224, 226, 235, 306 Yoga, 79-87
Viscous, 116, 117, 225, 226 YouTube, 85, 86, 138, 139, 141
Wales, 207 α- cut set, 355

Wave propagation, 116 β- cut set, 355
Weak convergence, 40 γ- cut set, 355
Also of Interest
By the same editors

DATA WRANGLING: Concepts, Applications, and Tools, edited by M.
Niranjanamurthy, Kavita Sheoran, Geetika Dhand, and Prabhjot Kaurk,
ISBN: 9781119879688. Written and edited by some of the world’s top
experts in the field, this exciting new volume provides state-of-the-art
research and latest technological breakthroughs in next-data wrangling, its
theoretical concepts, practical applications, and tools for solving everyday
problems.
ADVANCES IN DATA SCIENCE AND ANALYTICS, Edited by M.

Niranjanamurthy, Hemant Kumar Gianey, and Amir H. Gandomi, ISBN:
9781119791881. Presenting the concepts and advances of data science and
analytics, this volume, written and edited by a global team of experts, also
goes into the practical applications that can be utilized across multiple dis-
ciplines and industries, for both the engineer and the student, focusing on
machining learning, big data, business intelligence, and analytics.
WIRELESS COMMUNICATION SECURITY: Mobile and Network Security

Protocols, Edited by Manju Khari, Manisha Bharti, and M. Niranjanamurthy,
ISBN: 9781119777144. Presenting the concepts and advances of wireless
communication security, this volume, written and edited by a global team
of experts, also goes into the practical applications for the engineer, stu-
dent, and other industry professionals.
MEDICAL IMAGING, Edited by H. S. Sanjay, and M. Niranjanamurthy

ISBN: 9781119785392.Written and edited by a team of experts in the field,
this is the most comprehensive and up-to-date study of and reference for
the practical applications of medical imaging for engineers, scientists, stu-
dents, and medical professionals.
SECURITY ISSUES AND PRIVACY CONCERNS IN INDUSTRY 4.0
APPLICATIONS, Edited by Shibin David, R. S. Anand, V. Jeyakrishnan,
and M. Niranjanamurthy, ISBN: 9781119775621. Written and edited by a
team of international experts, this is the most comprehensive and up-to-
date coverage of the security and privacy issues surrounding Industry 4.0
applications, a must-have for any library.
Check out these other related titles from Scrivener

Publishing
CONVERGENCE OF DEEP LEARNING IN CYBER-IOT SYSTEMS AND

SECURITY, Edited by Rajdeep Chakraborty, Anupam Ghosh, Jyotsna
Kumar Mandal and S. Balamurugan, ISBN: 9781119857211. In-depth
analysis of Deep Learning-based cyber-IoT systems and security which
will be the industry leader for the next ten years.
MACHINE INTELLIGENCE, BIG DATA ANALYTICS, AND IOT IN

IMAGE PROCESSING: Practical Applications, Edited by Ashok Kumar,
Megha Bhushan, José A. Galindo, Lalit Garg and Yu-Chen Hu, ISBN:
9781119865049. Discusses both theoretical and practical aspects of how
to harness advanced technologies to develop practical applications such as
drone-based surveillance, smart transportation, healthcare, farming solu-
tions, and robotics used in automation.
MACHINE LEARNING TECHNIQUES AND ANALYTICS FOR

CLOUD SECURITY, Edited by Rajdeep Chakraborty, Anupam
Ghosh and Jyotsna Kumar Mandal, ISBN: 9781119762256. This
book covers new methods, surveys, case studies, and policy with
almost all machine learning techniques and analytics for cloud secu-
rity solutions.
WILEY END USER LICENSE AGREEMENT
Go to www.wiley.com/go/eula to access Wiley’s ebook EULA.

Ghosh S. Mathematics and Computer Science Vol 2. 2023

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ghosh S. Mathematics and Computer Science Vol 2. 2023

Uploaded by

Copyright:

Available Formats

Mathematics and Computer

Advances in Data Engineering and Machine Learning

Wiley Global Headquarters

Limit of Liability/Disclaimer of Warranty

Library of Congress Cataloging-in-Publication Data

Front cover images supplied by Wikimedia Commons

Printed in the USA

5 Novel Hybrid Optimal Deep Network and Optimization

7 Encoder-Decoder Models for Protein Secondary Structure

10 IoT Based Ensemble Predictive Techniques to Determine

12.3.2 Saddle Node Bifurcation in Genetic Toggle 180

14.4 Methodology 214

17.2 Assumptions and Notations 253

20 Automorphisms of Some Non-Abelian p−Groups of Order p4 291

23.4.4 Cost of the System 343

26.3.1 Theorem 2 389

Department of Computer Science, Aligarh Muslim University, Aligarh, UP, India

Keywords: Text mining, text classification, spam detection, text preprocessing,

*Corresponding author: ayaamer@myamu.ac.in

Features classification Predict test

Figure 1.1 Overview of text classification.

Table 1.1 Text classification compared model classifiers.

1.2 Text Mining Techniques

1.2.1 Data Mining

Classification TM Techniques Clustering

Information Summarization Information

Machine Learning Algorithms Knowledge

Figure 1.2 Text data mining techniques.

functionality is used for classification and prediction by using a set of pre-

1.2.2 Information Retrieval

from the greatest knowledge data sets. IR techniques are predominantly

1.2.3 Natural Language Processing (NLP)

1.2.4 Information Extraction

Figure 1.3 Information extraction.

distinguish the important relationships involved in the document as shown

1.2.5 Text Summarization

summarization extracts the important information or sentence from the

1.2.6 Text Categorization

1.2.8 Information Visualization

categorization, or importance of the tag. Information visualization or

1.2.9 Question Answer

key words extraction

User Interface Answering handel

Figure 1.4 Question answering system.

1.3 Dataset and Preprocessing Steps

1.3.1 Text Preprocess

1.4 Feature Extraction

Table 1.2 Preprocessing steps.

1.4.1 Term Frequency – Inverse Document Frequency

1.4.2 Bag of Words (BoW)

1.5 Supervised Machine Learning Classification

1.7 Experimentation and Discussion Results for

Table 1.4 Confusion matrix in machine learning.

Table 1.5 Spam dataset classification.

Spam Dataset classification

Figure 1.5 Spam dataset classification.

1.8 Text Mining Applications

1.9 Text Classification Support

Business Social Media

Figure 1.6 Applications of text data mining [32].

1.9.2 Business and Marketing

14. K. N. S. S. V Prasad, S. K. Saritha, and D. Saxena, “A Survey Paper on Concept

29. S. Alami and O. Elbeqqali, “Cybercrime profiling: Text mining techniques to

1. T (τ)- tumor cell population number (cells).

3. L (τ)- cytotoxic T lymphocytes (CD8+) concentration per