You are on page 1of 7

4

__________________________________________________________________

A Logical Approach to Fuzzy MADM Problems

4.1 Introduction
The classical MADM methods assume all measures of performance of alternatives
(mij) and weights of attributes (wj) values are crisp numbers. The alternatives with
higher overall or composite performance scores are considered to be preferred by
the decision maker. Since the final scores are real numbers, the preferred
alternatives are those with higher overall or composite performance scores. In
reality, measure of performance (mij) can be crisp, fuzzy and/or linguistic. For
example, let a material be chosen for making an engineering component and the
attributes considered are: cost of material, tensile strength, hardness, density, and
corrosion resistance. The last attribute, corrosion resistance, is not quantifiable;
rather, it is represented by linguistic terms such as ‘low’, ‘average’, ‘high’, etc. The
other attributes can be crisp numbers. This MADM problem contains a mixture of
fuzzy and crisp data. Most of the real-world MADM problems are of this type.
Fuzzy MADM methods are proposed to solve problems that involve fuzzy
data. Bellman and Zadeh (1970) were the first to relate fuzzy set theory to
decision-making problems. Yager and Basson (1975) proposed fuzzy sets for
decision making. Bass and Kwakernaak (1977) proposed a fuzzy MADM method
that is widely regarded as the classic work of fuzzy MADM methods. During the
last three decades, several fuzzy MADM methods have been proposed and
reviewed (Chen and Hwang, 1992; Triantaphyllou and Lin, 1996; Triantaphyllou,
2000; Figueira et al., 2004). After a systematic and critical study of the existing
fuzzy MADM methods, it has been found that the majority of the approaches
require cumbersome computations. As a result, none of them are suitable for
solving problems with more than 10 alternatives associated with more than 10
attributes. That drawback certainly limits their applicability to real-world problems.
Further, most approaches require that the elements in the decision matrix be
presented in a fuzzy format, though they are crisp in nature. Such an assumption
violates the original intent of fuzzy set theory. If the data is precisely known, there
is no subjectivity involved in the decision problem. Such data should never be
represented in any fuzzy format. The conversion of crisp data into fuzzy format
will increase the computational requirements. This, in turn, makes these fuzzy
44 Decision Making in the Manufacturing Environment

methods cumbersome to use, and incapable of solving problems that contain more
than 10 alternatives and 10 attributes.
Chen and Hwang (1992) proposed an approach to avoid the abovementioned
difficulties, so that MADM problems can be meaningfully and efficiently solved in
a fuzzy environment. The approach is composed of two phases. The first phase
converts fuzzy data into crisp scores. The result of the first phase is a decision
matrix that contains only crisp data. In the second phase, MADM methods,
described in Chapter 3, can be utilized to determine the ranking order of
alternatives. The easy-to-use and easy-to-understand characteristics of this
approach make it valuable to management and system analysts.

4.2 Method Proposed by Chen and Hwang (1992)


The method proposed by Chen and Hwang (1992) first converts linguistic terms
into fuzzy numbers and then the fuzzy numbers into crisp scores. The method is
described below.

4.2.1 Converting Linguistic Terms to Fuzzy Numbers

This method systematically converts linguistic terms into their corresponding fuzzy
numbers. It contains eight conversion scales. The conversion scales were proposed
by synthesizing and modifying the works of Wenstop (1976), Bass and
Kwakernaak (1977), Efstathiou and Rajkovic (1979), Bonissone (1982), Efstathiou
and Tong (1982), Kerre (1982), and Chen (1988),
To demonstrate the method, a 5-point scale having the linguistic terms low,
fairly low, medium, fairly high, and high, as shown in Figure 4.1 (Chen and
Hwang, 1992), is considered. These linguistic terms can be equated to other terms
like low, below average, average, above average, and high.

4.2.2 Converting Fuzzy Numbers to Crisp Scores

The method uses a fuzzy scoring approach that is a modification of the fuzzy
ranking approaches proposed by Jain (1976, 1977), and Chen (1985). The crisp
score of fuzzy number ‘M’ is obtained as follows:

µmax (x) = x, 0 ≤ x ≤ 1
0, otherwise (4.1)

µmin (x) = 1 - x, 0 ≤ x ≤ 1
0, otherwise (4.2)

The fuzzy max and fuzzy min of fuzzy numbers are defined in a manner such
that absolute locations of fuzzy numbers can be automatically incorporated in the
comparison cases. The left score of each fuzzy number ‘Mi’ is defined as
µL(Mi) = Sup[µmin(x) ^ µMi(x)] (4.3)
x
A Logical Approach 45

Figure 4.1. Linguistic terms to fuzzy numbers conversion (5-point scale) (from Chen and
Hwang 1992; with kind permission of Springer Science and Business Media)

The µL(Mi) score is a unique, crisp, real number in (0, 1). It is the maximum
membership value of the intersection of fuzzy number Mi and the fuzzy min. The
right score may be obtained in a similar manner:
µR (Mi) = Sup[µmax(x) ^ µMi(x)] (4.4)
x

Again, µR (Mi) is a crisp number [0,1]. Given the left and right scores, the total
score of a fuzzy number Mi is defined as:
µT (Mi) = [µR(Mi) + 1 - µL(Mi)] / 2 (4.5)

4.3 Demonstration of the Method


Now, the 5-point scale is considered to demonstrate the conversion of fuzzy
numbers into crisp scores (Figure 4.1).
Linguistic term Fuzzy number
Low M1
Below average M2
Average M3
Above average M4
High M5
The maximizing and minimizing sets are defined as Equations 4.1 and 4.2.
From Figure 4.1, membership functions of M1, M2, M3, M4, and M5 are written as:
46 Decision Making in the Manufacturing Environment

µM1(x) = 1, x = 0
(0.3-x) / (0.3), 0 ≤ x ≤ 0.3

µM2(x) = (x-0)/ (0.25), 0 ≤ x ≤ 0.3


(0.5-x) / (0.25), 0.25 ≤ x ≤ 0.5

µM3(x) = (x-0.3)/ (0.2), 0.3 ≤ x ≤ 0.5


(0.7-x)/ (0.2), 0.5 ≤ x ≤ 0.7

µM4(x) = (x-0.5)/ (0.25), 0.5 ≤ x ≤ 0.75


(1.0-x)/ (0.25), 0.75 ≤ x ≤ 1.0

µM5(x) = (x-0.7)/ (0.3), 0.7 ≤ x ≤ 1.0


1, x = 1

The right, left, and total scores are computed as follows for M1:

µR (M1) = Sup [µmax (x) ^ µM1(x)] = 0.23


x

µL(M1) = Sup [µmin (x) ^ µM1(x)] = 1.0


x

µT (M1) = [µR (M1) + 1 - µL(M1)] / 2 = 0.115

Similarly, the right, left, and total scores are computed for M2, M3, M4, and M5
and are tabulated as follows:

i µR (Mi) µL (Mi) µT (Mi)


1 0.23 1.0 0.115
2 0.39 0.8 0.295
3 0.58 0.59 0.495
4 0.79 0.4 0.695
5 1.0 0.23 0.895

Hence, the linguistic terms with their corresponding crisp scores are given in
Table 4.1. Instead of assigning arbitrary values for various attributes, this fuzzy
method reflects the exact linguistic descriptions in terms of crisp scores. Hence, it
gives better approximation of linguistic descriptions that are widely used.
It may be added here that this method can be used not only for assigning
values to the attributes, but also for deciding the relative importance between the
attributes. For example, using the same 5-point scale, the relative importance
between two attributes can be described as given in Table 4.2.
A Logical Approach 47

Table 4.1. Conversion of linguistic terms into fuzzy scores (5-point


scale)
__________________________________________________________
Linguistic term Fuzzy number Crisp score
__________________________________________________________
Low M1 0.115
Below average M2 0.295
Average M3 0.495
Above average M4 0.695
High M5 0.895
__________________________________________________________

Table 4.2. Conversion of linguistic terms into fuzzy scores (relative importance value on a
5-point scale)
________________________________________________________________________
Linguistic term Fuzzy number Crisp score
________________________________________________________________________
One attribute is very less important than the other M1 0.115
One attribute is less important than the other M2 0.295
Two attributes are equally important M3 0.495
One attribute is more important than the other M4 0.695
One attribute is much more important than the other M5 0.895
________________________________________________________________________

The decision makers can appropriately make use of any of the eight scales
suggested by Chen and Hwang (1992). For example, an 11-point scale is shown in
Figure 4.2, and the corresponding crisp scores of the fuzzy numbers are given in
Table 4.3.

Table 4.3. Conversion of linguistic terms into fuzzy scores (11-point


scale)
__________________________________________________________
Linguistic term Fuzzy number Crisp score
__________________________________________________________
Exceptionally low M1 0.045
Extremely low M2 0.135
Very low M3 0.255
Low M4 0.335
Below average M5 0.410
Average M6 0.500
Above average M7 0.590
High M8 0.665
Very high M9 0.745
Extremely high M10 0.865
Exceptionally high M11 0.955
__________________________________________________________
48 Decision Making in the Manufacturing Environment

Figure 4.2. Linguistic terms to fuzzy numbers conversion (11-point scale) (from Chen and
Hwang 1992; with kind permission of Springer Science and Business Media)

Using the same 11-point scale, the relative importance between two attributes
can be described as given in Table 4.4.

Table 4.4. Conversion of linguistic terms into fuzzy scores (relative importance value on an
11-point scale)
__________________________________________________________________________
Linguistic term Fuzzy number Crisp score
__________________________________________________________________________
One attribute is exceptionally less important than the other M1 0.045
One attribute is extremely less important than the other M2 0.135
One attribute is very less important than the other M3 0.255
One attribute is less important than the other M4 0.335
One attribute is slightly less important than the other M5 0.410
Two attributes are equally important than the other M6 0.500
One attribute is slightly more important than the other M7 0.590
One attribute is more important than the other M8 0.665
One attribute is much more important than the other M9 0.745
One attribute is extremely more important than the other M10 0.865
One attribute is exceptionally more important than the other M11 0.955
__________________________________________________________________________

It may be remembered that Tables 2.1 and 2.2 are suggested in Chapter 2 for
assigning the objective values to the subjective attributes, and for assigning the
relative importance between the attributes, respectively. Now, Tables 4.1 (or 4.3)
and 4.2 (or 4.4) may be used for the same purpose, as these give better
A Logical Approach 49

approximation of the linguistic terms. The case studies presented in Chapters 5–30
of this book utilize Tables 4.3 and 4.4.

References
Bass SJ, Kwakernaak H (1977) Rating and ranking of multi-aspects alternatives
using fuzzy sets. Automatica 13:47–58
Bellman RE, Zadeh LE (1970) Decision-making in a fuzzy environment.
Management Science 17:212–223
Bonissone PP (1982) A fuzzy sets based linguistic approach: theory and
applications. In: Gupta MM, Sanchez E (eds) Approximate reasoning in
decision analysis. North Holland, pp 329–339
Chen SH (1985) Ranking fuzzy numbers with maximizing set and minimizing set.
Fuzzy Sets and Systems 17:113–129
Chen SM (1988) A new approach to handling fuzzy decision-making problems. In:
Proc. 18th International Symposium on Multiple-Valued Logic, Computer
Society Press, Palma de Mallorca, Spain
Chen SJ, Hwang CL (1992) Fuzzy multiple attribute decision making-methods and
applications. Lecture Notes in Economics and Mathematical Systems,
Springer, New York
Efstathiou J, Rajkovic V (1979) Multiattribute decision making using a fuzzy
hewristic approach. IEEE Transactions on Systems, Man, Cybernetics, SMC-
9:326–333
Efstathiou J, Tong R (1980) Ranking fuzzy sets using linguistic preference
relations. In: Proc. 10th International Symposium on Multiple-Valued Logic,
Northwestern University, Evanston
Figueira J, Greco S, Ehrgott M (2004) Multiple criteria decision analysis: state of
the art surveys. Springer, New York
Jain R (1976) Decision making in the presence of fuzzy variables. IEEE
Transactions on Systems, Man and Cybernetics, SMC 6:698–703
Jain R (1977) A procedure for multi-aspect decision making using fuzzy sets.
International Journal of System Science 8:1–7
Kerre EE (1982) The use of fuzzy set theory in electrocardiological diagnostics. In:
Gupta MM, Sanchez E (eds) Approximate reasoning in decision analysis.
North Holland, pp 277–282
Triantaphyllou E (2000) Multi-criteria decision making methods: a comparative
study. Kluwer Academic Publishers, Dordrecht
Triantaphyllou E, Lin CT (1996) Development and evaluation of five fuzzy multi-
attribute decision making methods. International Journal of Approximate
Reasoning 14:281–310
Wenstop F (1976) Fuzzy set simulation models in a systems dynamic perspective.
Kybernetes 6:209–218
Yager RR, Basson D (1975) Decision-making with fuzzy sets. Decision Sciences
6:590–600

You might also like