Review and Applications of MCDM Techniques and Development of A Methodology For Selection of The Most Suitable DM Technique

You might also like

Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 18

Review and Applications of MCDM Techniques

and Development of a Methodology for Selection


of the Most Suitable DM Technique

Akshat Jain (2014UME1003)


Prachanda Poudel (2014UME1016)
Priyank Tada (2014UME1495)
Ridham Patoliya (2014UME21285)

Guided by,
Dr. A.P.S. Rathore

Dept. of Mechanical
Engineering, 2017-2018
1. Introduction
In today’s world, all types of organizations are complex, dynamic, and uncertain.
Organizations are continuously exposed to a myriad of challenges, including innovation,
technological disruptions, global competition, leadership change, and shifting economic,
social, and regulatory conditions. To ensure survival and growth, decision makers must lead
the organizations with an effective data-driven transformation strategy. To achieve the
goals of their organizations, managers must understand how decisions are made and know
which decision-making tools to use. To a great extent, the success or failure of both people
and organizations depends on the quality of their decisions.

To make a rational decision, one must use the collection of facts and information to
come up with a structured solution. By the type of information and facts that are gathered,
decision making techniques are generally divided in main two categories: Qualitative
technique and Quantitative technique. In this study, we have included only quantitative
techniques for multicriteria decision making. We have considered time span of 2008 to
2018 (Last decade) for our study.

Multi-Criteria Decision Making (MCDM) techniques are generally used when there
are many criteria governing the process of decision making. MCDM methods have been
designed to designate a preferred alternative, classify alternatives in a small number of
categories, and/or rank alternatives in a subjective preference order( ref 1).
Characteristics of MCDM: (ref 1, ref 3)

 MCDM methods deal with discrete alternatives, which are described by a set of
criteria. ref 1
 Criteria values can be determined as a cardinal or ordinal information. ref 1
 An MCDM problem can be easily expressed in matrix format. ref 1
 Data collection is a crucial part of any MCDM and inconsistency in data leads to
practically incorrect decision.
 Information could be determined exactly or could be fuzzy, determined in intervals.
ref 1
 Most of the MCDM techniques require that the attributes be assigned weights of
importance(ref 3).
 Even after completing MCDM analysis, decision makers have to assess the quality of
the final results.

2. History of Decision Making


From the philosophical era of Plato and Confucius to the time of artificial intelligence and
advanced computing, decision making has always been considered as a pivotal process in
any field. Philosophers ponder what our decisions say about our values; historians dissect
the choices leaders make at critical junctures. Research into risk and organizational
behavior springs from a more practical desire: to help managers achieve better outcomes.

For millennia, human decisions are guided by interpretations of entrails, smoke, dreams,
and the like; In ancient times the most mundane happenings were connected with the
grandest cosmic events. Indian ancient literature dictates patterns of stars as driving force
for everyday events happening around us. But these abstract concepts can’t be applied in
the present dynamically complex systems in organizations.

Industrial revolution has altered the way we see organizational complexity. Fragmentation
of work at all operational levels requires decisions to be holistic in a way that they involve
all the factors affecting outcomes of the decisions. To make good decisions we need to
have concrete data and that is why quantitative techniques for decision making has evolved
as a major research field.
MCDM techniques and their years of development can be summarized though this table

MCDM Technique Year of Major Development


Analytic Hierarchy Process 1980
Analytic Network Process 1996
ELECTRE 1996
TOPSIS 1981
VIKOR 1998
DEMATEL 1976
DEA 1978
PROMETHEE 1984
CPORAS 1994
SAW 1968

3. Research Objectives
 To Find out which processes are most widely used in the field of decision making in
operations research.
 To extract major types of decisions
 To study comparative analysis of different MCDM techniques
 To develop a method for selection of an MCDM technique for a specific decision-
making problem
 To apply an appropriate technique in a real-life decision-making problem justifying
the methodology used for selection of an MCDM.
4. Research Methodology
Study of evolution of processes
for Decision Making

Review of research articles


based on MCDM techniques
(2008-2018)

Shortlisting of top 5 MCDM


techniques based on the extent
of use in the past decade

Temporal statistical analysis of


the extent of use of MCDM
techniques

Intensive study of the top 5


techniques

Developing a methodology for


selection of an MCDM

Application of a suitable
technique in a real-life problem
5. Statistical Analysis of Extent of use of
MCDM techniques
For statistical analysis of extent of use of MCDM techniques, we performed “keyword
search” on two authentic research databases; ScienceDirect and Emerald Insights. We
searched research articles related to the processes in the time span of 2008 to 2018. Total
number of research articles are shown in the last column. Top five processes were decided
according to popularity in research trend during 2008 to 2018.

MCDM Technique Science Emerald Total


Direct Insight
Analytic Hierarchy Process (AHP) 1551 269 1820

Technique for Order Preference by Similarity to Ideal 250 148 398


Solution (TOPSIS)
Analytic Network Process (ANP) 340 75 415

Decision Making Trial and Evaluation Laboratory 113 40 153


(DEMATEL)
VIKOR 200 23 223

Preference Ranking Organization Method for 4 3 7


Enrichment of Evaluations (PROMETHEE)
Elimination Et Choice Translating Reality(ELECTRE) 6 0 6

Compromise Programming (CP) 74 2 76

Multi-attribute utility theory (MAUT) 63 7 70

Simple Additive Weighting (SAW) 66 3 69

Weighted sum model (WSM) 15 17 32

Multi-attribute value theory (MAVT) 16 1 17

Weighted product model (WPM) 5 9 14

Measuring Attractiveness by a Categorical Based 8 1 9


Evaluation Technique (MACBETH)

Utility Additive (UTA) 1 8 9

Additive Ratio Assessment (ARAS) 7 0 7

Complex Proportional Assessment of Alternatives 6 1 7


(COPRAS)
Pie Chart Analysis for MCDM techniques

AHP

6%
8%
Technique for Order Preference by
Similarity to Ideal Solution (TOPSIS)
11%
47%
ANP

Decision Making Trial and Evaluation


28% Laboratory (DEMATEL)

VIKOR
Temporal Statistical Analysis for Extent of use of MCDM
Techniques
This method of temporal statistical analysis has been used to visualize how research trends
for different MCDM varies with time. Numbers of research articles published in each year
were tabulated for all the processes. The table below shows numbers of research articles
published and their respective years. Here, since the sole purpose was to study the
research trend, only one research database (sciencedirect.com) was considered.

2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
AHP 87 91 84 126 129 133 159 182 204 244
ANP 18 28 37 39 39 27 33 45 29 24
TOPSIS 18 44 39 81 76 85 94 126 149 171
VIKOR 1 5 7 17 15 22 17 31 24 32
DEMATEL 3 7 12 31 33 21 18 32 24 37

244

204

182
171
159
149

129 133
126 126

91 94
87 84 85
81
76

44 45
39
37 39 39 37
31 33 33 32
31 29 32
28 27 24 24
18 22
21 18
17 15 17
12
3 7
5 7
1
2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

AHP ANP TOPSIS VIKOR DEMATEL


6. Brief introduction of top 5 techniques

AHP Method of Decision Making

AHP stands for “Analytic Hierarchy Process”. This method is used to determine the
weightage of the criteria that we consider for decision making between the available
alternatives. This process is very useful when we have intangible criteria for our decision
making process as it takes absolute as well as ordinal data as input. The matrices between
Alternative v/s criteria and criteria v/s criteria will be formed and criteria weightage and
alternative ranking can be obtained when necessary calculation is done. The drawback of
this process is that it does not take the feedback into consideration for decision making.

ANP Method of Decision Making


The Analytic Network Process (ANP) is a generalization of the Analytic Hierarchy Process
(AHP). The basic structure is an influence network of clusters and nodes contained within
the clusters. Priorities are established in the same way they are in the AHP using pairwise
comparisons and judgment. Many decision problems cannot be structured hierarchically
because they involve the interaction and dependence of higher-level elements in a
hierarchy on lower-level elements. Not only does the importance of the criteria determine
the importance of the alternatives as in a hierarchy, but also the importance of the
alternatives themselves determines the importance of the criteria. Feedback enables us to
factor the future into the present to determine what we have to do to attain a desired
future.

TOPSIS Method of Decision Making

TOPSIS stands for “Technique for Order Performance by Similarity to Ideal Solution”. It is
mostly used for multiple attribute decision making. In this method we construct two
solutions, most ideal solution and worst solution which are denoted by x+ and x-. This
process sets the two benchmark i.e. of being close to the most ideal solution and being far
from the worst solution. It makes use of the Absolute values for decision making. This is one
of the drawback of the TOPSIS method as we can’t obtain the solution if absolute data is
not available. However taking the absolute data in use, the most reliable decision making
solution for multi-attribute decision making can be obtained by using this process.
DEMATEL Method of Decision Making
DEMATEL stands for “Decision Making Trail and Evaluation Laboratory”. This method is
widely used as a reliable method of decision making to solve the cause and effect among
the evaluation criteria. In this process, the opinions of the experts are gathered, and Z
matrix is calculated. On normalizing this matrix we get D matrix, which will be later used in
calculating a T matrix. Now the sum of row and column of this matrix is enlisted as an Eigen
vector. Then we move forward and determine the Threshold value α, which is sum average
of the whole matrix. Now the cause and effect diagram will be made and if the initial cause
and effect diagram is acceptable then final cause and effect diagram is made. This will be
our solution for decision making.

VIKOR Method of Decision Making


VIKOR stands for “VlseKriterijumska Optimizacija I Kompromisno Resenje”. This method is
used to rank the alternatives Ai as per the values of scalar quantities Maximum Group
Utility (Si), Minimum Individual Regret of the opponent (Ri) and (Qi), to be calculated for
each alternative. To proceed with this process, two values i.e. best and the worst
Alternatives for each criteria is determined in the matrix formed. After this the scalar
quantities are calculated. The value of Qi gives us the ranking of the alternative as the
alternative with minimum Q value is most favorable and the alternative with maximum Q
value being most undesired.
Comparative analysis of top 5 MCDM
After reviewing around 200 research articles for each process, we came to the final table
for extent of use of top 5 MCDM. These are empirical results which have to be justified
using criteria for selection of suitable MCDM.

To validate the feasibility and robustness of a MCDM method, sensitivity analysis


should be conducted. To evaluate the MCDM methodologies, we defined
6subcriteria such as Mathematical logical procedure; Applicability to intangibles;
Justification of the approach and its axioms; Validation of the decision process in the
prioritization of both tangibles and intangibles; Prediction of outcomes with tangibles
and intangibles; Generality of the process to decisions with dependence and
feedback; Applicability to conflict resolution as a decision process and finally,
trustworthiness and validity of the approach.

Of all the available criteria, ranking is given to the process as high, medium or low.
This is done depending upon various factors for each criterion as follows:

Simplicity of execution: A method is rated low if its logic is complicated and can
only be used by the professional decision makers or experts; medium if one needs
to put much efforts to learn it; high if it can be easily understood and implemented by
most users in practice. Hence VIKOR and TOPSIS being the easiest to apply are
given High ranking and AHP, ANP and DEMATEL have been given Medium.

Comprehensive structure: Breadth and Depth: structure is considered to be


broad if there are many independent distinct elements (criteria and subcriteria), and
be broader if it contains loops and feedback. A structure is deep if each element is
further decomposed to the most detailed elements.
A method is rated low if it contains only a few criteria without decomposition to sub-
criteria, sub-sub-criteria with the necessary detail; medium if it’s structure is neither
broad nor deep; high if it’s structure is not only broad (contains a number of distinct
criteria) but also deep (criteria can be broken down to sub-criteria). Here, ANP
makes use of feedback from sub criteria level, it is given a high rating while others
are given Medium ranking.

Logical, Mathematical procedure: Mathematical logical procedure means there is


a formal mathematical representation of the logic and reasoning behind a theory.
A method is rated low if it involves only a simple mathematical logical procedure
(e.g. arithmetic value utility); medium if it uses references sequence or relative
difference to rank alternatives; high if it uses a pairwise comparison technique to
determine the dominance of one criterion over another. Here, VIKOR and TOPSIS
doesn’t make use of pair wise comparison and are given Medium ranking.

Scales of measurement: Scales of measurement are used to categorize or


quantify the decision variables.
A method is rated low if it uses the nominal or ordinal scale; medium if it uses
interval scales or ratio scales; high if it uses absolute scales.

Ranking of tangibles: To aid decision makers make a decision, ranking makes it


possible to evaluate complex information according to certain criteria and produce
the total preorder of alternatives by balancing the preferences on all the criteria.
A method is rated NA if it does not involve some kind of ranking of the alternatives;
low if uses a nominal scale; medium if it uses an ordinal scale; high if it uses a
cardinal scale to rank alternatives. Since we are using ordinal scale for decision
making in all the tools we give everyone Medium ranking.

Sensitivity analysis: A method is rated low if it involves only a single parameter;


medium if it works on two to three parameters; high if it is capable of assessing
more than three parameters. Since all five processes can assess more than three
parameters, they all are given high ranking.

Generalizability to dependence and feedback: Feedback enables the decision


maker to factor the future into the present to determine what he must do to attain a
desired future. Also feedback can cause an unimportant element to become
important (Saaty 2004).
A method is rated NA if it does not involve feedback in the process of making a
decision; low if either dependence or feedback is used indirectly, medium if both
partially influence the results of decision making; high if they are regarded as parts
of the inputs of the decision itself that can affect the results of that decision. In
accordance to this ANP is given High rating and the rest are Low in rating.

Applicability to conflict resolution: It means that the method should provide a


way for each party in the conflict to evaluate the costs and the benefits to both sides
to determine “fair” tradeoffs by giving up some of what it has, in return for getting
what it wants from the other party (Peniwati 2007).
A method is rated low if it uses a simple mathematical compensation technique to
make tradeoffs in a conflict; medium if it uses an analytical method for dealing with
conflict resolution; high if it can provide the best solution for a group conflict that is
understandable, acceptable, practical and flexible.

Trustworthiness and validity of the approach: Trustworthiness has been defined


as that quality of a method and its findings that make it noteworthy to decision
makers.
A method is rated low if it uses ordinal measurement model with a simple structure;
medium if it uses cardinal measurement model but does not provide rigorous
mathematical axioms; high if it uses a cardinal measurement model with
mathematical logical procedure and mathematical axioms. Other than VIKOR all the
processes have proper mathematical logical procedure and axioms.

Limitations of top 5 MCDM


Practical Application of MCDM in an Medium
Scale Enterprise
Problem Statement: A brass part manufacturing firm, which currently purchases casted raw
materials from supplier, wants to set up a casting plant of its own. There are two options
available for the investor to choose from: 1. Manual casting Process, 2. Extrusion casting
process. The main criteria influencing the decision are related to Design, Quality,
Production and cost. There are five sub criteria under design: weight, length of the section,
thickness of the section, minimum core size and shape complexity. There are three sub-
criteria which influence quality of casting: surface roughness, density variation along
section, impurity. Five criteria influences production features of the process: Production
rate, Product variety, Lead time, setup time, material utilization. Cost criteria is influenced
by these five factors: Tooling cost, direct labour cost, equipment cost, finishing cost,
electricity cost.

Casting process

Quality Production Design Cost

Surface
Production rate Weight Tooling cost
roughness

Density variation Length of the Direct labour


Product variety
along section section cost

Thichness of the
Impurity Lead time Equipment cost
section

Minimum core
Setup time Finishing cost
size

Material Shape
Electricity cost
utilization complexity
Selection of decision making techniques:

The type of decision to be taken is selection. For selection type of decision making problem, from table __
it is recommended to use AHP or ANP or VIKOR. From comparative analysis of MCDM techniques done in
this course of research, we came to a conclusion that AHP is the most suitable technique for this decision.
Other two processes can be later applied to validate the comparative analysis. Currently, we have
implemented AHP technique.

Collection of data:

We interviewed 4 experts who are business partners and also fully conversant with the alternatives and
criteria. They were told to give importance rating according to classical AHP scale developed by Saaty.

Implementation of AHP:
Pair-wise comparison matrix for main criteria of the problem

Table 1 Design Quality Production Cost Weightage


Design 1.0000 0.1250 0.5000 0.2000 0.0547
Quality 8.0000 1.0000 7.0000 4.0000 0.6045
Production 2.0000 0.1429 1.0000 0.2000 0.0826
Cost 5.0000 0.2500 5.0000 1.0000 0.2582
Consistency Ratio calculation:

λmax = 4.1771
n = number of criteria

C.I. = (λmax – n)/(n-1) = 0.059

R.I.= 0.9

C.R. = C.I / R.I. = 0.066

Since C.R. < 0.1, the data is consistent and we can go for further data processing for AHP

Pair-wise comparison for sub-criteria under Design criteria:

Design Weight Length of the Thickness of Minimum core Shape Weightage


section section size complexity
Weight 1.0000 0.1429 0.2000 0.2000 0.1429 0.0372
Length of 7.0000 1.0000 1.0000 1.0000 0.1429 0.1373
the section
Thickness of 5.0000 1.0000 1.0000 1.0000 0.1429 0.1213
section
Minimum 5.0000 1.0000 1.0000 1.0000 0.2000 0.1283
core size
Shape 7.0000 7.0000 7.0000 5.0000 1.0000 0.5760
complexity

All the sub-criteria care evaluated against two available alternatives: (Example of shape complexity is
given below)

Table 5 (Shape Manual casting Extrusion casting process


complexity) process
Manual casting 1.0000 0.1429 0.1250
process
Extrusion casting 7.0000 1.0000 0.8750
process

Global priorities for alternatives:

Manual Casting Process 0.1175


Extrusion Casting Process 0.8825

Results: Alternative 2, Extrusion Casting Process is more suitable for the firm.
Remaining Work:
- Application of VIKOR and ANP for the same problem
- Sensitivity analysis of all processes

You might also like