9th IEEE International Conference on Communication Systems and Network Technologies

Hand Written Digit Recognition using Machine Learning

Rohan Sethi1, Ila Kaushik2*
1
Department of Computer Science and Engineering
2
Department of Information Technology
1
H.M.R Institute of Technology & Management, Delhi, India
2
Krishna Institute of Engineering & Technology, Ghaziabad, India
rohan_sethi-22.98@ieee.org, ila.kaushik.8.10@gmail.com
Abstract— Hand-written character and digit recognition have as the printed text or hand-written text involves conversion
been one of the most exigent and engrossing field of pattern process which is aimed for generation of machine generated
recognition and image processing. The main aim of this paper and machine recognisable text due which it is space
is to demonstrate and represent the work which is related to efficient.[6]
hand-written digit recognition. The hand-written digit The technique is so, well developed that it has various
recognition is a very exigent task. In this recognition task, the applications, like as it is efficiently utilised by the visionless
numbers are not accurately written or scripted as they differ in and visually impaired, also used by banks to digitize
shape or size; due to which the feature extraction and documents.[7]
segmentation of hand-written numerical script is arduous. The
Also, the ‘Barcode Recognition’ technique has a closely
vertical and horizontal projections methods are used for the
linked foundation with OCR technique; Moreover, used in
purpose of segmentation in the proposed work. SVM is applied
for recognition and classification, while Convex hull algorithm ‘Number Plate Recognition’, education, finance and other
is applied for feature extraction. fields.[8]
The various factors which are responsible for the
Keywords-KNN, OCR, Naïve Bayes, Neural Networks, difference in the hand writing such as multiple orientations,
Feature extraction, Optical character recognition, Segmentation, skewness of the lines, pressure points of the connected
Handwritten digit classification. components, overlapping characters etc. [9]
Every character in the language could be expressed in
I. INTRODUCTION many different forms, which depends on hand-writing of an
individual, which offers difficulty to recognise the digit
Optimal Character Recognition (OCR), can be defined as accurately.[10]
the method which is capable to recognise and identify the The six steps in Optical Digit Recognition are as follows:
text by the conversion of typed images, hand-written or the
text which is in the printed form into machine-encoded text.
The machine-encoded text is in the form of ASCII or
Unicode.[1]
The pixel representation of a letter is obtained after the
conversion from equivalent character or digit representation.
The process is a simulation by the machine of human
reading. Digit recognition can be stated as the research of
how observation of the environment by the machines is able
to distinguish the digit of interest and predicts the accurate
decision related to the digits and characters.[2]
OCR, (Optical Character Reader) can be defined as the Figure 1. Optical Digit/Character Recognition Steps
device which is capable of translating the images into
machine recognisable formats which could be in the form of
ASCII or Unicode.[3] The process of character or digit x Image Acquisition: Images are obtained by training
recognition of text which is hand-written is more difficult data-set or by involving the scan of the
than text generated by machines, as the digits in hand-written document.[11]
text are non-uniform and due to which it is variable in shape x Pre-Processing: The process or technique which is
and size; while the machine printed text are simple and applied to remove the noise from the image and
uniformly spaced which are aligned fixed.[4] leads to the enhancement of the image for further
The process contributes in the automation process, as it processing. [12] These techniques are applied only
increases the necessary interaction among the machines and after image acquisition and segmentation. The steps
humans.[5] In OCR, there are two types of OCR Datasets: involved in this technique are, skew correction, noise
Offline and Online. In ‘Offline’ dataset, the datasets are removal, normalization, cropping and resizing,
loaded and provided to the classifier or training model as an binarization, thinning, and skeletonization.
input, but in ‘Online’ dataset, the co-ordinates of the
x Segmentation: Segmentation can be defined as the
detection of drawing the digit is recorded to instantiate the process, which divides an entity into parts or
prediction. The OCR technique involves numerous benefits sections. It is of 3 types, line, word and character. As

978-1-7281-4976-9/20/$31.00 ©2020 IEEE 49

DOI: 10.1109/CSNT.2020.

Authorized licensed use limited to: University of Melbourne. Downloaded on June 23,2020 at 19:11:12 UTC from IEEE Xplore. Restrictions apply.
 the terms are self-explanatory, line segmentation
involves division with respect to lines, character
segmentation involves division with respect to
characters and word segmentation involves division
with respect to words. But, here in this paper we
have implemented a different approach to
segmentation; we have involved splitting of dataset
during the process of segmentation. And, according
to splitting process the prediction of the labelled
digit is more accurate, as processed by the
classifier.[13]
x Feature Extraction: The process of ‘Feature
Extraction’ can be defined as the process in which
the essential information about the focussed subject
is extracted which is existing in the image.
x Classification: Within classification, the pre-defined
class is assigned with an unknown sample. As the
features, are extracted the digits are classified and
recognised accordingly.
x Post-Processing: The techniques of post processing
are used to achieve more accuracy.
II. LITERATURE SURVEY
The hand-written digit classification or recognition, for
the cursive hand-written document, the study demonstrated
that off-line hand-written analysis of the document occurs
through skew recognition, writing pressure detection and
segmentation. The segmentation method which was
proposed, was based on modified vertical and horizontal
projections; moreover, in the existence of multi-skewed and
over-lapped text lines these projections are capable to
segment the text lines and the words accurately.[14] The
testing of the method was executed on more than 550 images
of text which were of IAM database and sample images of
handwriting of different writers on different back-grounds.
This proposed method was capable of performing correct
segmentation of around 92.56% words and 95.65% lines
from the IAM dataset. Moreover, around 96% lines are
perfectly normalized with a minute error correction. The
method of skew normalization demonstrates the skew angle
accurately and compares it to different hands on techniques
efficiently [15].
The segmentation of lines, of the text is processed on the
basis of information energy targeted individually for the
calculation of every pixel and “Artificial Neural Network”
recognised the characters and digits. The accuracy of
recognition was around 92% [16].
The execution of feature set which is hull based and of
convex form which implies, 125 features to be computed
according to the consideration of diverse attributes of bays of
a pattern of the convex hull, isolated Bangla basic characters
and digits were also recognised and this technique was also
included in this field of study. The accuracy rate of
recognition of hand-written Bangla characters was 76.86%
and for Bangla numerals it was 99.45% [17].
In this field, unprecedented technique is applied for the
recognition of Bangla cursive words. For, the recognition
purpose MLP classifier is used.[18] The categorization of
50
Authorized licensed use limited to: University of Melbourne. Downloaded on June 23,2020 at 19:11:12 UTC from IEEE Xplore. Restrictions apply.
words at the initial stage of recognition analysis actually is
carried out in three parts i.e. upper, middle and lower parts.
The features which are diversified, are designed after
considering octant-centroid features, shadow features which
are modified and features, which are in the long -run. The
accuracy rate of 80.58% was recorded when the experiments
were carried out on 300 samples [19].
The study of different segmentation techniques was
included in hand-written character recognition. There are
three levels of segmentation process discussed, which
includes character, word and text-lines; the needs and the
factors which leads to any affect to the segmentation process
are discussed.
The study which involves the work in this field
demonstrates entirely a fresh approach which targets the use
of series of algorithms for recognition and segmentation for
the OCR of handwritten script and digits. Hidden Markov
Model (HMM) is processed with precision rate of 92.3% for
recognition with lexicon size 50.[20] The word-level
segmentation is derived from the combination of Lexicon
and HMM [21].
The discussion of various segmentation levels has been
existent in this field of work study.
‘Hough Transformation’ can be defined as the level used
for segmentation of text lines. The skeletonization process
occurs due to the fact that vertically connected components
are separated. The experiments which were implemented
were on IC-DAR2007 dataset.
The strength function of connectivity involves the
process of segmentation in this field of study.
“Connectivity strength parameter” can be defined as the
parameter which decides the constituents of the text-line.
This approach is adaptive to the language in nature with
precision rate of around 97.3%.[22]
III. SUPERVISED MACHINE LEARNING SYSTEM
The supervised machine learning technique can be
defined as the technique , in which certain pre-defined input
or pre-defined labels are given to the machine learning
model which acts as the supervisor to the machine learning
model and trains the model , such that the model could either
classify or regress with experience from the set of inputs
given to obtain the desired set of outputs .
Figure 2. Multi Variate Scattered Data
This technique performs on the basis of set of input-
output pairs, which associates the regression and
classification techniques to provide with the desired outputs.
As in this algorithm the model is trained first, with labelled
dataset the output can be predicted much earlier by the
humans; but, so as to associate the scenario with machines ,
supervised machine learning is used. The set of inputs given
to the model are generally known as Vectors and the desired
outputs obtained are known as Supervisory Signals.[23]
Since, this technique leads to the development of relation
between input and output, the mathematical formula
associated with this technique can be expressed as follows: -
………..(1)
The supervised machine learning algorithms can be
classified as: -
x Regression : The statistical supervised machine
learning algorithm, which determines the statistical
relationship between a single dependent variable
and multiple independent variables which is widely
used in prediction and financial analysis. The main
target of this analysis is to predict the best set of
values of the random variable or independent
variable on the foundation of dependent
variable.[24] The two types of regression analysis
are : Linear and Multiple regression.
x Classification : The classification as, statistical
supervised machine learning algorithm can be
explained as the technique or the approach which
expects a set of inputs to learn, observe and analyse
the data , and implement the experience to classify
on the set of new inputs after the model has been
trained well. The main motive of this technique is
to classify the data according to the set of features
persistent with the input data , after the successful
training of the model with the labelled dataset.
In this research paper, we have demonstrated and
executed the supervised machine learning algorithm of
classification so as to accomplish our purpose of
implementing the project of “Hand-written digit
classification”.
IV. DIVERSE CLASSIFICATION ALGORITHMS IN MACHINE
LEARNING
In statistics and machine learning, the classification
algorithms analyses the training dataset for the prediction of
g dataset.
target or testing
Figure 3. Classification Algorithm
The 7 types of classification algorithms are explained as
follows: -
x Naïve Bayes Classifier: The classification
algorithm which is mathematically based on
Bayes’ Theorem and assumes the independence of
predictors is known Naïve Bayes Classifier. This
classification technique on the basis of assumption
as stated above, the particular feature of a class is
non-related to the presence of the other feature.
51
Authorized licensed use limited to: University of Melbourne. Downloaded on June 23,2020 at 19:11:12 UTC from IEEE Xplore. Restrictions apply.
These properties contribute to the probability
independently. It is useful for large datasets as it is
not very much difficult to build.[25]
x Nearest Neighbour: The K-Nearest Neighbour
supervised machine learning algorithm can be
defined as algorithm in which labelled dataset is
taken as the input and the computation as desired
for the output, is held over. The feature space
contains the K-Nearest Neighbour training
examples. Also, it is an instance-based learning.
x Logistic Regression: The statistical technique
which analyses the dataset to determine the
outcome where there are one or more independent
variables. The outcome for this classification
technique is computed with the help of a
dichotomous variable, which is a variable with two
possible outcomes only. The main motive of this
algorithm is to search for a perfect fitting model
which depicts the relationship among the
dichotomous variables and independent
variables.[26]
x Decision Trees: The classification algorithm which
builds the regression and classification models in
the form of a structure of a tree. The associated tree
structure is formed by the sequential or incremental
division of the dataset in the subsets and further
subsets in more smaller subsets. The topmost
decision node of the tree is known as the root node,
which is the best predictor node of the decision
tree. It could handle numerical as well as
categorical data.
x Random Forest: The construction of multitude
decision trees at the training time which actually
formulates the decision trees for the combinational
learning methods for regression and classification.
They tend to overfit their training dataset for
correct decision trees.
x Neural Network: The neural network classification
algorithm in supervised machine learning contains
the units of neural network better known as neurons
which are known to convert the input vector into
considerable output and structurally organized in
layers. In this algorithm every unit expects an input
and applies the non-linear function on the input and
transfers the output thus, obtained to the next layer
for the computation. The necessary weights are
applied on to the signals so as to make the neural
network adaptable which have been turned in the
training phase. The main drawback of this technique
is that it does not support feedback system for every
neuron involved in the network
In this paper, we have applied the KNN Classification
Algorithm to support the implementation of our project
work, “Hand-written digit classification”.
 V. KNN – BASED ARCHITECTURE
K-Nearest Neighbors Algorithm, can be defined as the
supervised machine learning algorithm which is non-
parametric in nature and applied in the areas of classification
and regression. Considering both the cases, and on the
analysis of feature space the input comprises of the K closest
training examples. It is decided by the output that whether,
the algorithm will be used for regression or classification.
In the case of classification, the output obtains the
membership of the class.[27] The classification of the object
is purely based on the vote of its neighbors. Whereas, in
regression the property value for the object is output.
KNN, the supervised non-parametric classification or
regression-based algorithm which is also known by the term
lazy learning or late learning involves all the computations to
g of classification.[28]
be derived at the final stage
Figure 4. KNN Algorithm with training datasets
This classification algorithm involves local
approximation technique as it is instance-based learning
algorithm. In this algorithm the implementation is not too
difficult as there is a lack of generalized training phase.
…..(2)
Applications: The main areas of application of this
supervised machine learning algorithm are data
compression, genetic analysis and many more.
And, the eminent reason for implementing this algorithm is
when the data is abundant enough for perfect
implementation of this algorithm to execute the
classification as required. [29] Moreover, the direct solution
is obtained only when there exist non-linear decision
boundaries. The input features are characterised as, both
quantitative and qualitative. And, the output features are
categorical in nature. This algorithm explains the majority
votes of K nearest neighbours categorically, where the value
of K could differ. Significant contributions include,
parameter selection, presence of noise, feature selection and
curse of dimensionality. [30] The distance functions which
are applicable in KNN algorithm are: Euclidean distance,
Manhattan distance, Minkowski function, Hamming
function and Mahalanobis function.
Assumption
¾ The data assumption is near to none, as the
algorithm does not involve any assumption.
52
Authorized licensed use limited to: University of Melbourne. Downloaded on June 23,2020 at 19:11:12 UTC from IEEE Xplore. Restrictions apply.
¾ Selection of K parameter based on data.
¾ Distance metric is required to define proximity
among any two data points.
Steps for KNN Algorithm
x Computation of distance metric between testing data
points and all the labelled data points.
x The data points which are labelled are enjoined in the
ascending order of distance metric.
x Selection of the top K labelled data points.
x The class labels are matched with the K labelled data
points and assigned to the test data points.
VI. IMPLEMENTATION
Here we propose, KNN algorithm to solve the problem of
Hand-Written Digit Classification; the dataset used to solve
the problem is referred to as the MNIST dataset.
MNIST Dataset: Modified National Institute of Standards
and Technology dataset, can be defined as the large dataset
which is used to train various machine learning models. The
black and white images obtained from this dataset are
normalized in 28x28 pixel bounding box which are anti-
aliased and led to the introduction of grayscale levels. It
consists of 10,000 testing images and 60,000 training
images.
Figure 5. MNIST Dataset.
Problem Statement:
Given: A set of greyscale isolated numerical images
sourced from MNIST database.
The objective includes: -
9 To recognize handwritten digits accurately.
9 To improve the accuracy rate of detection.
9 To develop a method which is unconventional of
the digit-size and independent of the style of
writer/ink.
Scope of Study
• Postal mail sorting
• Courtesy amounts on cheques
• Formation of data entry etc.
The approach to find a solution to the problem of
handwritten numeral recognition is broadly classified into
three blocks:
i) Pre‐Processing
ii) Feature Extraction using PCA
iii) Classification using 1‐Nearest Neighbour
algorithm;
 The block diagram for the system is explained as
follows:-
Figure 6. System Diagram of the implementation of pattern classifier
Initially, around 5000 random images are taken from the
dataset, to train the machine learning model, then the dataset
is split into training and testing datasets. The segmentation of
dataset in training and testing, leads to the application of
KNN algorithm on the training dataset to perfectly analyse
the testing dataset images and classify them accordingly.
KNN algorithm, then computes the nearest equivalent label
to identify the best possible match of the hand-written digit
obtained as the source from the testing dataset by using the
Euclidean distance formula , to demonstrate the majority
vote of the dataset and then computing the output after the
thorough analysis and comparison between the class labels
and input image data. The accuracy is further calculated, for
the process. And, hence the images are classified on the basis
of KNN algorithm where in this paper we proposed the value
of k=5, according to our project.
The overall design related to the classification of digit
database of MNIST is explained in the following algorithm.
x Algorithm: Input Digits Classification, Numeral
images of MNIST Database which are in isolation
x Output: Recognition Technique of Numbers,
Features of the Structure and KNN classifier.
Algorithm:
Step 1: Consider a grey level image and convert it into a binary
image
Step 2: Pre-process the Binary Image
Step 3: Consider the Binary Image and convert it into a single
Dimensional Array [1, n]
Step 4: The label of every individual array must be kept along
with it.
Step 5: The classifier is fed with the training data-set.
Step 6: For all the images in the Sample and Test Database the
steps from 1 to 5 are repeated
Step 7: Estimate the minimum distance between feature vector
and vector stored in the library by using Euclidian distances.
Step 8: Feed the classifier with testing data set.
Step 9: The loaded images are classified into relevant class
labels using least distance classifier of K-nearest neighbour.
Step10: End the process.
(a)
53
Authorized licensed use limited to: University of Melbourne. Downloaded on June 23,2020 at 19:11:12 UTC from IEEE Xplore. Restrictions apply.
(b)
(c)
Figure 7. (a),(b),(c) Hand-Written Digit Classification
The images obtained after successfully training the
machine learning model, applying KNN algorithm and
Euclidean distance formula to compute the nearest and most
accurate labelled data so as to classify the digits correctly
from the testing dataset. The output obtained is in the form of
28x28 matrix, is in accordance with the labelled dataset and
training dataset after the completion of training the model in
supervised machine learning algorithm and the system
predicts the values of digits correctly.
VII. RESULT & DISCUSSION
The training of images for classification was achieved
with the assistance of KNN Algorithm. The number of K
nearest points to the labelled dataset, and in comparison,
with the training dataset the prediction by the training model
was perfectly achieved as desired. This algorithm was used
to classify the hand-written digits with accuracy and less
amount of computation. The two MNIST datasets used are:
Training and Testing; and the distance formula used with
assistance to KNN algorithm was Euclidean distance
formula.
VIII. CONCLUSION
The work deals with hand-written digit classification,
using the efficient KNN supervised machine learning
algorithm. For, training supervised classification machine
learning model the training dataset is fed to the classifier as
input with labelled data and the model after successful
training, becomes capable for the classification of hand-
written digit on the basis of any testing point fed to the
classifier, after the involvement of Euclidean distance which
determines the nearest labelled datapoint related to the
testing data point. For classification, KNN supervised
machine learning algorithm was used. And, for the
computation of best fit supervisory signal to the input vector
Euclidean distance formula was used. All the experiments
were implemented or executed in Jupyter Notebook, a web
application which runs on local machine server and is a part
of collection of applications in Anaconda software. The
datasets used were training and testing MNIST datasets.
 This work can be proliferated in future to result in rapid
computation which would also decrease time and would
increase efficiency and target better results.
REFERENCES
[1] Pritam Dhande, Reena Kharat. "Recognition of cursive English
handwritten characters" 2017 International Conference on Trends in
Electronics and Informatics (ICEI), 2017.
[2] Mandal*, S Shahnawazuddin, Rohit Sinha*, S. R. Mahadeva
Prasanna *, Suresh Sundaram*,“ Exploring Sparse Representation for
Improved Online Handwriting Recognition” 2018 16th International
Conference on Frontiers in Handwriting Recognition , 2018.
[3] Khamparia, A., & Singh, K. M. (2019). A systematic review on deep
learning architectures and applications. Expert Systems, 36(3). doi:
10.1111/exsy.12400
[4] Soni, S., & Bhushan, B. (2019). Use of Machine Learning algorithms
for designing efficient cyber security solutions. 2019 2nd
International Conference on Intelligent Computing, Instrumentation
and Control Technologies (ICICICT). doi:
10.1109/icicict46008.2019.8993253
[5] Manchanda, C., Rathi, R., & Sharma, N. (2019). Traffic Density
Investigation & Road Accident Analysis in India using Deep
Learning. 2019 International Conference on Computing,
Communication, and Intelligent Systems (ICCCIS). doi:
10.1109/icccis48478.2019.8974528
[6] Yuanzhi Zhu , Zecheng Xie , Lianwen Jin, Xiaoxue Chen, Yaoxiong
Huang and Ming Zhang, “SCUT-EPT : New Dataset and
Benchmark for Offline Chinese Text Recognition in Examination
Paper” IEEE Access , 2018.
[7] S.E Benita Galaxy , S. Selvin Ebenzer, “Enhancement of
segmentation and zoning to improve the accuracy of handwritten
character recognition” , International Conference on Electrical ,
Electronics and Optimization Techniques (ICEEOT) – 2016.
[8] Khamparia, A., Saini, G., Gupta, D., Khanna, A., Tiwari, S., &
Albuquerque, V. H. C. D. (2019). Seasonal Crops Disease Prediction
and Classification Using Deep Convolutional Encoder
Network. Circuits, Systems, and Signal Processing, 39(2), 818–836.
doi: 10.1007/s00034-019-01041-0
[9] Sil, R., Roy, A., Bhushan, B., & Mazumdar, A. (2019). Artificial
Intelligence and Machine Learning based Legal Application: The
State-of-the-Art and Future Research Trends. 2019 International
Conference on Computing, Communication, and Intelligent Systems
(ICCCIS). doi: 10.1109/icccis48478.2019.8974479
[10] Sugata Das , Sekhar Mandal, “Keyword Spotting in Historical Bangla
Handwritten Document Image Using CNN” , Second International
Conference on Advanced Computational and Communication
Paradigms (ICACCP) , 2019.
[11] Adeel Yousaf , M. Jaleed Khan , M. Imran , Khurram Khurshid,
“Benchmark Dataset for Offline Handwritten Character Recognition”
, IEEE Xplore , 2017
[12] Grover, M., Verma, B., Sharma, N., & Kaushik, I. (2019). Traffic
control using V-2-V Based Method using Reinforcement
Learning. 2019 International Conference on Computing,
Communication, and Intelligent Systems (ICCCIS). doi:
10.1109/icccis48478.2019.8974540
[13] Momina Moetesum*, Imran Siddiqi* , Chawki Djeddi , Yaacoub
Hannad , Somaya Al-Maadeed , “Data Driven Feature Extraction for
Gender Classification using Multi-script Handwritten Texts” , 16th
International Conference on Frontiers in Handwriting Recognition,
2018.
[14] Reza Tavoli , Mohammadreza Keyvanpour , “A method for
handwritten word spotting based on particle swarm optimisation and
multi-layer perceptron” , IET (The Institution of Engineering and
Technology) Journals , 2018.
[15] Hamayun A. Khan , “MCS HOG Features and SVM Based
Handwritten Digit Recognition System” , Scientific Research , 2017.
[16] Irfan Ali , Insaf Ali , Subhash , Asif Khan , Syed Ahmed Raza , Basit
Hassan , Priha Bhatti, “Sindhi Handwritten- Digits Recognition Using
Machine Learning Techniques” , International Journal of Computer
Science and Network Security (IJCSNS) , VOL.19 No.5 , 2019.
[17] Huseyin Kusetogullari , Amir Yavariabdi , Abbas Cheddad , Hakan
Grahn and Johan Hall, “ A Swedish Historical Handwritten Digit
Dataset” , Springer , 2019.
[18] Khamparia, A., Gupta, D., Albuquerque, V. H. C. D., Sangaiah, A.
K., & Jhaveri, R. H. (2020). Internet of health things-driven deep
learning system for detection and classification of cervical cells using
transfer learning. The Journal of Supercomputing. doi:
10.1007/s11227-020-03159-4
[19] Yuxiang Wang , Ruijin Wang , Dongfen Li , Daniel Adu-Gyamfi ,
Kaibin Tian and Yixin Zhu, “Improved Handwritten Digit
Recognition using Quantum K-Nearest Neighbor Algorithm”
Springer , 2019.
[20] Tiwari, R., Sharma, N., Kaushik, I., Tiwari, A., & Bhushan, B.
(2019). Evolution of IoT & Data Analytics using Deep
Learning. 2019 International Conference on Computing,
Communication, and Intelligent Systems (ICCCIS). doi:
10.1109/icccis48478.2019.8974481
[21] Mahreen Ahmed , Asma Ghulam Rasool , Hammad Afzal , Imran
Siddiqi , “Improving handwriting based gender classification using
ensemble classifiers” , ScienceDirect , 2017.
[22] Khamparia, A., Singh, A., Anand, D., Gupta, D., Khanna, A., Kumar,
N. A., & Tan, J. (2018). A novel deep learning-based multi-model
ensemble method for the prediction of neuromuscular
disorders. Neural Computing and Applications. doi: 10.1007/s00521-
018-3896-0
[23] Emad Sami Jaha , “Efficient Gabor-Based Recognition for
Handwritten Arabic-Indic Digits” , International Journal of Advanced
Computer Science and Applications Vol . 10 No. 1 , 2019.
[24] Hui-huang Zhao and Han Liu , “Multiple classifiers fusion and CNN
feature extraction for handwritten digits recognition” , Springer ,
2019.
[25] U. Chauhan, V. Kumar, V. Chauhan, S. Tiwary and A. Kumar,
"Cardiac Arrest Prediction using Machine Learning Algorithms,"
2019 2nd International Conference on Intelligent Computing,
Instrumentation and Control Technologies (ICICICT),
Kannur,Kerala, India, 2019, pp. 886-890.
[26] Saqib Ali, Zeeshan Shaukat, Muhammad Azeem, Zareen Sakhawat,
Tariq Mahmood and Khali ur Rehman, “An efficient and improved
scheme for handwritten digit recognition based on convolutional
neural network” , Springer , 2019.
[27] Nicole Dalia Cilia, Claudio De Stefano , Francesco Fontanella,
Alessandra Scotto di Freca, “A ranking-based feature selection
approach for handwritten character recognition” ScienceDirect ,
2019.
[28] Jindal, M., Gupta, J., & Bhushan, B. (2019). Machine learning
methods for IoT and their Future Applications. 2019 International
Conference on Computing, Communication, and Intelligent Systems
(ICCCIS). doi: 10.1109/icccis48478.2019.8974551
[29] Khamparia, A., Pandey, B., Pandey, D. K., Gupta, D., Khanna, A., &
Albuquerque, V. H. C. D. (2020). Comparison of RSM, ANN and
Fuzzy Logic for extraction of Oleonolic Acid from Ocimum
sanctum. Computers in Industry, 117, 103200. doi:
10.1016/j.compind.2020.103200
[30] Harjani, M., Grover, M., Sharma, N., & Kaushik, I. (2019). Analysis
of Various Machine Learning Algorithm for Cardiac Pulse
Prediction. 2019 International Conference on Computing,
Communication, and Intelligent Systems (ICCCIS). doi:
10.1109/icccis48478.2019.8974519

54

Authorized licensed use limited to: University of Melbourne. Downloaded on June 23,2020 at 19:11:12 UTC from IEEE Xplore. Restrictions apply.