Vous êtes sur la page 1sur 19

Fuzzy group decision-making

for facility location selection


Cengiz Kahraman
a,
*
, Da Ruan
b
, Ibrahim Doggan
a
a
Department of Industrial Engineering, Istanbul Technical University,
80680 Macka, Istanbul, Turkey
b
Belgian Nuclear Research Centre (SCK-CEN), Boeretang 200, B-2400 Mol, Belgium
Received 10 March 2002; received in revised form 14 February 2003; accepted 17 March 2003
Abstract
The selection of a facility location among alternative locations is a multicriteria
decision-making problem including both quantitative and qualitative criteria. The
conventional approaches to facility location problem tend to be less eective in dealing
with the imprecise or vagueness nature of the linguistic assessment. Under many situ-
ations, the values of the qualitative criteria are often imprecisely dened for the deci-
sion-makers. The aim of the paper is to solve facility location problems using dierent
solution approaches of fuzzy multi-attribute group decision-making. The paper includes
four dierent fuzzy multi-attribute group decision-making approaches. The rst one is a
fuzzy model of group decision proposed by Blin. The second is the fuzzy synthetic
evaluation. The third is Yagers weighted goals method and the last one is fuzzy analytic
hierarchy process. Although four approaches have the same objective of selecting the
best facility location alternative, they come from dierent theoretic backgrounds and
relate dierently to the discipline of multi-attribute group decision-making. These ap-
proaches are extended to select the best facility location alternative by taking into ac-
count quantitative and qualitative criteria. A short comparative analysis among the
approaches and a numeric example to each approach are given.
2003 Published by Elsevier Inc.
Keywords: Facility location; Fuzzy sets; Synthetic evaluation; Weighted goals; Group
decision; AHP
*
Corresponding author. Tel.: +90-212-293-1300/2073; fax: +90-212-240-7260.
E-mail address: kahramanc@itu.edu.tr (C. Kahraman).
0020-0255/$ - see front matter 2003 Published by Elsevier Inc.
doi:10.1016/S0020-0255(03)00183-X
Information Sciences 157 (2003) 135153
www.elsevier.com/locate/ins
1. Introduction
Evaluation of alternative regions, sub-regions, and communities is com-
monly termed macroanalysis. Evaluation of specic sites in the selected com-
munity is commonly termed microanalysis. Techniques used to support
macroanalysis include factor-rating systems, linear programming, center of
gravity, and analytic Delphi model [17].
The other conventional approaches to location selection include heuristics
[5,31], integer programming [37], dynamic programming [12], nonlinear pro-
gramming [39], multi-objective goal programming [2], nonconvex program-
ming [11], quadratic programming [20], analog approach [1], analytic hierarchy
process [2], multi-attribute utility method [4], multiple regression analysis [40],
and branch and bound method [12]. These approaches can only provide a set of
systematic steps for problem solving without considering the relationships
between the decisions factors globally. Moreover, the ability and experience of
the analyst(s) may also inuence signicantly the nal outcome. In addition,
articial intelligence (AI) techniques, such as expert systems, articial neural
networks (ANNs), and fuzzy set theory are used in location selection.
Recent research shows that the applications of ANN techniques in decision-
making domain are very promising. Wang and Malakooti [49] present a feed-
forward neural network using the golden section descent technique for multiple
criteria decision-making. Benjamin et al. [4] compare the performance of
ANNs as classiers in the facility location domain. Adaptive resonance theory
(ART II) and back propagation (BP) paradigms are used as exemplars of
ANNs developed using supervised and unsupervised learning. Jungthirapanich
[25] develops a decision support system incorporated a database management
system (DBMS), a linear additive multi-attribute utility method, an expert
system, and graphical support. The DBMS stores location data collected
from public documents or reports. The multi-attribute utility method is used
to generate a suitability index for each location. The expert system is developed
to identify the top few locations and explain their advantages and disadvan-
tages.
The organization of this paper is as follows. First, the literature review on
fuzzy facility location selection is given. Second, the issues in facility location
are introduced. These are the tangible and intangible criteria of facility location
problem. Then four dierent fuzzy multi-attribute group decision-making ap-
proaches are used to select the best facility location. The rst one is a fuzzy
model of group decision proposed by Blin [8]. The second is the fuzzy synthetic
evaluation. The third one is Yagers [50] weighted goals method and the last
one is fuzzy analytic hierarchy process (AHP). These approaches are ex-
tended to select the best facility location alternative by taking into account
both intangible and tangible criteria. A numeric example is given for each
approach.
136 C. Kahraman et al. / Information Sciences 157 (2003) 135153
2. Literature review: fuzzy sets approaches to facility location selection
Humans are unsuccessful in making quantitative predictions, whereas they
are comparatively ecient in qualitative forecasting. Further, humans are
more prone to interference from biasing tendencies if they are forced to provide
numerical estimates since the elicitation of numerical estimates forces an in-
dividual to operate in a mode which requires more mental eort than that
required for less precise verbal statements [29]. Since fuzzy linguistic models
permit the translation of verbal expressions into numerical ones, thereby
dealing quantitatively with imprecision in the expression of the importance of
each criterion, some multi-attribute methods based on fuzzy relations are used.
Applications of fuzzy sets within the eld of decision-making have, for the
most part, consisted of extensions or fuzzications of the classical theories of
decision-making. While decision-making under conditions of risk and uncer-
tainty have been modeled by probabilistic decision theories and by game the-
ories, fuzzy decision theories attempt to deal with the vagueness or fuzziness
inherent in subjective or imprecise determinations of preferences, constraints,
and goals [51].
In the facility location problem, the conventional approaches tend to be less
eective in dealing with the imprecision or vagueness nature of the linguistic
assessment. There has been an increasing interest for fuzzy sets to be used for
the facility location problem in the recent years.
Tzeng and Chen [47] propose a location model based on a fuzzy multi-
objective approach. The model helps in determining the optimal number and
sites of re stations at an international airport, and also assists the relevant
authorities in drawing up optimal locations for re stations. Because of the
combinatorial complexity of their model, a genetic algorithm is employed and
compared with the enumeration method. Kuo et al. [33] develop a decision
support system using the fuzzy sets theory being integrated with analytic hi-
erarchy process for locating a new convenience store. Chen [18,19] proposes a
new multiple criteria decision-making method to solve the distribution center
location selection problem under fuzzy environment. In the proposed method,
the ratings of each alternative and the weight of criterion are described by
linguistics variables that can be expressed in triangular fuzzy numbers. The
nal evaluation value of each distribution center location is also expressed in a
triangular fuzzy number. By calculating the dierence of nal evaluation value
between each pair of distribution center locations, a fuzzy preference relation
matrix is constructed to represent the intensity of the preferences of one plant
location over another. And then, a stepwise ranking procedure is proposed to
determine the ranking order of all candidate locations. Kuo et al. [32] develop a
decision support system for locating convenience store. The proposed system
consists of four components: (1) hierarchical structure development for fuzzy
analytic hierarchy process (fuzzy AHP), (2) weights determination, (3) data
C. Kahraman et al. / Information Sciences 157 (2003) 135153 137
collection, and (4) decision-making. In the rst component, reviewing the related
references and interviewing the retailing experts formulate the hierarchical
structure of fuzzy AHP. Then, a questionnaire survey is conducted to determine
the weight of each factor in the second component, while the corresponding
data are collected through some government publications and actual investi-
gation. Finally, a feed-forward neural network with error back-propagation
learning algorithm is applied to study the relationship between the factors and
the store performance.
3. Fuzzy group decision-making for facility location selection
When decisions which are made by more than one person are modeled, two
dierences from the case of a single decision-maker can be considered: rst, the
goals of the individual decision-makers may dier such that each places a
dierent ordering on the alternatives; second, the individual decision-makers
may have access to dierent information upon which to base their decision.
Theories known as n-person game theories deal with both of these consider-
ations, team theories of decision-making deal only with the second, and group-
decision theories deal only with the rst.
Kacprzyk et al. [26] present new concepts of solutions in group decision-
making under fuzzy individual preference relations and a fuzzy majority ex-
pressed by a fuzzy linguistic quantier (e.g., most, almost all, . . .). They use two
fuzzy-logic-based calculi of linguistically quantied proposition, underlying
Zadehs [52] representation of commonsense knowledge as a collection of
dispositions (propositions with implicit linguistic quantiers).
Bezdek et al. [6,7], Blin [8], Blin and Whinston [9], Montero and Tejada [38],
Nurmi [41], and Tanino [45,46] interpret group decision-making as follows:
Suppose you have a set of n options, S = (s
1
; s
2
; . . . ; s
n
), and m individuals.
Each individual k, k = 1; . . . ; m, provides his or her preferences over S. As these
preferences may be not clear-cut, their representation by individual fuzzy
preference relations is strongly advocated.
Lee [34] builds a group decision-making structure model of risk in software
development and proposes two algorithms to tackle the rate of aggregative risk
in a fuzzy environment by fuzzy sets theory during any phase of the life cycle.
Hsu and Chen [24] propose a method for aggregating individual fuzzy opinions
into a group fuzzy consensus opinion. He presents a procedure for aggregating
the expert opinions. Herrera et al. [21] present a consensus model in group
decision-making under linguistic assessments. It is based on the use of linguistic
preferences to provide individuals opinions, and on the use of fuzzy majority
of consensus, represented by means of a linguistic quantier. Herrera et al. [22]
present several group decision-making processes by a direct approach. These
processes are designed using the linguistic ordered weighted averaging
138 C. Kahraman et al. / Information Sciences 157 (2003) 135153
(LOWA) operator. Herrera et al. [23] propose a model for the consensus
reaching problem in heterogeneous group decision-making. This model con-
tains two types of linguistic consensus measures: linguistic consensus degrees
and linguistic proximities to guide the consensus reaching process. Li [36] in-
vestigates the problems of decision-making with multiple judge, multiple cri-
teria in a fuzzy environment, where the performance of alternatives and the
importance of criteria are imprecisely dened and represented by fuzzy sets. A
fuzzy model associated with the solution algorithm is proposed on the basis of
an a-level weighted, fuzzy preference relation. Bender and Simonovic [3] de-
scribe a fuzzy compromise approach to decision analysis within the context of
water resource systems planning under uncertainty. The fuzzy compromise
approach allows a family of possible conditions to be reviewed, and supports
group decisions through fuzzy sets designed to reect collective opinions and
conicting judgments. Chen [18,19] presents a new algorithm to evaluate the
rate of aggregative risk in software development using fuzzy set theory under
the group decision-making environment.
3.1. Blins fuzzy relations
Fuzzy relations map elements of one universe, say X, to those of another
universe, say Y , through the Cartesian product of the two universes. The
strength of the relation between ordered pairs of the two universes is not
measured with the characteristic function, but rather with a membership
function expressing various degrees of strength of the relation on the unit
interval [0; 1[. Hence a fuzzy relation

RR is a mapping from the Cartesian space
X Y to the interval [0; 1[, where the strength of the mapping is expressed by
the membership function of the relation for ordered pairs from the two uni-
verses, or l
~
RR
(x; y) [42].
A fuzzy model of group decision proposed by Blin [8] is extended to select a
facility location. Each member of a group of n individual decision-makers is
assumed to have a reexive, anti-symmetric, and transitive preference ordering
P
k
, k
n
, which totally or partially orders a set X of alternatives. A social
choice function is then found which, given the individual preference ordering.
Basically, the model allows for the individual decision-makers to possess dif-
ferent aims and values while still assuming that the overall purpose is to reach a
common acceptable decision. In order to deal with the multiplicity of opinion
evidenced in the group, the social preference S may be dened as a fuzzy binary
relation with membership grade function [30]:
l
S
: X X [0; 1[ (1)
which assigns the membership grade l
S
(x
i
; x
j
) indicating the degree of group
preference of alternative x
j
over alternative x
j
.
C. Kahraman et al. / Information Sciences 157 (2003) 135153 139
The expression of the group preference requires some appropriate means of
aggregating the individual preferences. One simple method computes the rela-
tive popularity of alternative x
i
over x
j
by dividing the number of persons pre-
ferring x
i
to x
j
, denoted by N(x
i
; x
j
), by the total number of decision-makers, n:
l
S
(x
i
; x
j
) =
N(x
i
; x
j
)
n
(2)
Once the fuzzy relationship S has been dened, the nal non-fuzzy group
preference can be determined by converting S into its resolution form
S =
_
a
aS
a
(3)
which is the union of the crisp relations S
a
comprising the a-cuts of the fuzzy
relation S, a A
S
(the level set of S), each scaled by a. Each value a essentially
represents the level of agreement between the individuals concerning the par-
ticular crisp ordering S
a
. One procedure that maximizes the nal agreement
level consists of intersecting the classes of crisp total orderings that are com-
patible with the pairs in the a-cuts S
a
for increasingly smaller values of a until a
single crisp total ordering is achieved. In this process, any pairs (x
i
; x
j
) that lead
to an intransitivity are removed. The largest value a for which the unique
compatible ordering on X X is found represents the maximized agreement
level of the group and the crisp ordering itself represents the group decision
[30].
3.1.1. A numeric example to Blins fuzzy relations for the selection among facility
location alternatives
The ve main attributes are considered and for each attribute, every group
member makes a preference ordering among the facility location alternative.
The attributes considered are political risk, government barriers, other facili-
ties, business climate. Then all the preference orderings are combined by a
procedure that maximizes the nal agreement level. The procedure consists of
intersecting the classes of crisp total orderings that are compatible with pairs
in the a-cuts S
a
for increasingly smaller values of a until a single crisp total
ordering is achieved.
Each individual of a group of eight decision-makers has a total preference
ordering P
i
(i N
8
) on a set of alternatives, where FLA is facility location
alternative, X = FLA
I
; FLA
II
; FLA
III
; FLA
IV
as follows [28]:
P
1
= FLA
I
; FLA
II
; FLA
III
; FLA
IV

P
2
= P
5
= FLA
IV
; FLA
III
; FLA
II
; FLA
I

P
3
= P
7
= FLA
II
; FLA
I
; FLA
III
; FLA
IV

P
4
= P
8
= FLA
I
; FLA
IV
; FLA
II
; FLA
III

P
6
= FLA
IV
; FLA
I
; FLA
II
; FLA
III

140 C. Kahraman et al. / Information Sciences 157 (2003) 135153


Using the membership function given in Eq. (2) for the fuzzy group prefer-
ence ordering relation S (where n = 8), the following fuzzy social preference
relation is obtained. First, the relative popularities of an FLA to another one
are given:
l
S
(FLA
I
; FLA
III
) =
6
8
= 0:75 l
S
(FLA
II
; FLA
III
) =
6
8
= 0:75
l
S
(FLA
I
; FLA
IV
) =
5
8
= 0:625 l
S
(FLA
III
; FLA
I
) =
2
8
= 0:25
l
S
(FLA
II
; FLA
I
) =
4
8
= 0:5 l
S
(FLA
II
; FLA
IV
) =
3
8
= 0:375
l
S
(FLA
I
; FLA
II
) =
4
8
= 0:5 l
S
(FLA
III
; FLA
II
) =
2
8
= 0:25
l
S
(FLA
III
; FLA
IV
) =
3
8
= 0:375 l
S
(FLA
IV
; FLA
I
) =
3
8
= 0:375
l
S
(FLA
IV
; FLA
II
) =
5
8
= 0:625 l
S
(FLA
IV
; FLA
III
) =
5
8
= 0:625
The relative popularities obtained are summarized in Table 1.
The a-cuts of this fuzzy relation S are
S
1
= O
S
0:75
= (FLA
I
; FLA
III
); (FLA
II
; FLA
III
)
S
0:625
= (FLA
I
; FLA
IV
); (FLA
IV
; FLA
II
); (FLA
IV
; FLA
III
);
(FLA
I
; FLA
III
); (FLA
II
; FLA
III
)
S
0:5
= (FLA
II
; FLA
I
); (FLA
I
; FLA
II
); (FLA
I
; FLA
IV
); (FLA
IV
; FLA
II
);
(FLA
IV
; FLA
III
); (FLA
I
; FLA
III
); (FLA
II
; FLA
III
)
S
0:375
= (FLA
IV
; FLA
I
); (FLA
II
; FLA
IV
); (FLA
III
; FLA
IV
);
(FLA
II
; FLA
I
); (FLA
I
; FLA
II
); (FLA
I
; FLA
IV
); (FLA
IV
; FLA
II
);
(FLA
IV
; FLA
III
); (FLA
I
; FLA
III
); (FLA
II
; FLA
III
)
S
0:25
= (FLA
III
; FLA
I
); (FLA
III
; FLA
II
); (FLA
IV
; FLA
I
);
(FLA
II
; FLA
IV
); (FLA
III
; FLA
IV
); (FLA
II
; FLA
I
);
(FLA
I
; FLA
II
); (FLA
I
; FLA
IV
); (FLA
IV
; FLA
II
);
(FLA
IV
; FLA
III
); (FLA
I
; FLA
III
); (FLA
II
; FLA
III
)
Now the procedure to arrive at the unique crisp ordering which constitutes the
group choice can be applied. All total orderings on X X are, of course,
Table 1
Fuzzy social preference relations
I II III IV
I 0.0 0.5 0.75 0.625
II 0.5 0.0 0.75 0.375
III 0.25 0.25 0.0 0.375
IV 0.375 0.625 0.625 0.0
C. Kahraman et al. / Information Sciences 157 (2003) 135153 141
compatible with the empty set of S
1
. The total orderings O
0:75
that are com-
patible with the pairs in the crisp relation S
0:75
are
O
0:75
= (FLA
IV
; FLA
I
; FLA
II
; FLA
III
); (FLA
I
; FLA
II
; FLA
III
; FLA
IV
);
(FLA
I
; FLA
IV
; FLA
II
; FLA
III
); (FLA
I
; FLA
II
; FLA
IV
; FLA
III
);
(FLA
IV
; FLA
II
; FLA
I
; FLA
III
); (FLA
II
; FLA
I
; FLA
III
; FLA
IV
);
(FLA
II
; FLA
IV
; FLA
I
; FLA
III
); (FLA
II
; FLA
I
; FLA
IV
; FLA
III
)
Thus,
O
1
O
0:75
= O
0:75
The orderings compatible with S
0:625
are
O
0:625
= (FLA
I
; FLA
IV
; FLA
II
; FLA
III
)
and
O
1
O
0:75
O
0:625
= (FLA
I
; FLA
IV
; FLA
II
; FLA
III
)
Thus, the value 0.625 represents the group level of agreement concerning the
social choice denoted by the total ordering (FLA
I
; FLA
IV
; FLA
II
; FLA
III
).
3.2. Fuzzy synthetic evaluation
An important application of the fuzzy transform used in developing Zadehs
extension principle is synthetic evaluation. The term synthetic is used to con-
note the process of evaluation whereby several individual elements and com-
ponents of an evaluation are synthesized into an aggregate form; the whole is a
synthesis of the parts. The key here is that the various elements can be numeric
or non-numeric, and the process of fuzzy synthesis is naturally accommodated
using synthetic evaluation. In reality, an evaluation of an object, especially an
ill-dened one, is often vague and ambiguous. The evaluation is usually de-
scribed in natural language terms, since a numerical evaluation is often too
complex, too unacceptable, and too ephemeral (transient) [42].
Let X be a universe of factors and Y be a universe of evaluations, so
X = x
1
; x
2
; . . . ; x
n
and Y = y
1
; y
2
; . . . ; y
m
. Let

RR = [r
ij
[ be a fuzzy relation
where i = 1; 2; . . . ; n and j = 1; 2; . . . ; m. Let ~ ww be a fuzzy vector of scoring
factors. The sum of the scores is unity. So we have
~ ww = w
1
; w
2
; . . . ; w
n
where

i
w
i
= 1 (4)
At this stage, to determine the relative importance of the criteria, Saatys an-
alytic hierarchy process can be used [43]. Fig. 1 shows part of a questionnaire
form, which might facilitate the answering of pairwise comparison questions.
After obtaining the pairwise judgments, the next step is the computation of a
vector of priorities or weighting of elements in the matrix. In terms of matrix
142 C. Kahraman et al. / Information Sciences 157 (2003) 135153
algebra, this consists of calculating the principal vector (eigenvector) of the
matrix, and then normalizing it to sum to 1.0.
After a group of decision-makers lled in the questionnaire, the geometric
mean of the preference numbers is used to obtain a single preference number
[44]. Then the elements of each column is divided by the sum of that column
(i.e., normalize the column) and the elements in each resulting row are added
and this sum is divided by the number of elements in the row.
The process of determining a grade for a specic alternative is equivalent to
the process of determining a membership value for the alternative in each of
the evaluation categories, y
i
. This process is implemented through the com-
position operation,
~ee = ~ ww

RR (5)
where ~ee is a fuzzy vector containing the membership values for the alternative
in each of the y
i
evaluation categories.
3.2.1. A numeric example of fuzzy synthetic evaluation for the selection among
FLAs
Suppose we want to compare two FLAs. The evaluation criteria are prox-
imity to customers (PC), infrastructure (I), quality of labor (QL), free trade
zones (FTZ), and competitive advantage (CA). FLAs are measured against
these criteria and given ratings categorized as excellent (E), superior (S), ade-
quate (A), and inferior (I). Excellent means that FLA is the best available
with respect to the particular criterion. Superior means that FLA is among
the best with respect to this criterion. Adequate means that, although not
superior, FLA can meet the minimum acceptable requirements for this crite-
rion. Inferior means that FLA cannot meet the requirements for the par-
ticular criterion. Suppose that the relations based on the consensus of the
group of decision-makers are as in the Table 2.
With respect to: Best
overall FLA:



Criteria:
9

A
b
s
o
l
u
t
e
8

7

V
e
r
y

s
t
r
o
n
g

6

5

F
a
i
r
l
y

s
t
r
o
n
g

4

3

W
e
a
k

2

1

E
q
u
a
l

2

W
e
a
k

3

4

5

F
a
i
r
l
y

s
t
r
o
n
g

6

7

V
e
r
y

s
t
r
o
n
g
8

9

A
b
s
o
l
u
t
e
C
r
i
t
e
r
i
a

A
A
A
B
C
D
B
B
C
D
C D
Fig. 1. Questionnaire form, which could be used to facilitate preference comparisons.
C. Kahraman et al. / Information Sciences 157 (2003) 135153 143
Now the criterion vector, ~ ww, will be obtained by using Saatys analytic hi-
erarchy process. The matrix of relative weights of subjective estimates are given
in Table 3.
Using the matrix of subjective attribute weights in Table 3 and Saatys
method, the following vector is obtained:
w
aw
= (0:537; 0:230; 0:100; 0:100; 0:033)
Now the composition ~ee is calculated as 0:537; 0:400; 0:300; 0:200 for FLA
I
and 0:200; 0:500; 0:400; 0:300 for FLA
II
. FLA
I
has its highest membership in
the category excellent and FLA
II
has its highest membership in the category
superior. Hence, FLA
II
should be selected.
3.3. Yagers weighted goals method
Let X = x
1
; x
2
; . . . ; x
n
be a set of alternatives. The goals are represented by
the fuzzy sets

GG
j
, j = 1; 2; . . . ; m. The importance (weight) of goal j is ex-
pressed by w
j
. The attainment of goal

GG
j
by alternative x
i
is expressed by the
degree of membership l
~
GG
j
.
The decision is dened as the intersection of all fuzzy goals, that is,

DD =

GG
w
1
1


GG
w
2
2


GG
w
m
m
(6)
Table 2
Fuzzy relations between alternatives and evaluation criteria
Criteria Alternatives
FLA
I
FLA
II
E S A I E S A I
PC 0.7 0.4 0.3 0.1 0.2 0.5 0.4 0.3
I 0.5 0.6 0.3 0.2 0.2 0.5 0.8 0.4
QL 0.2 0.4 0.7 0.2 0.3 0.6 0.4 0.2
FTZ 0.8 0.7 0.2 0.1 0.7 0.8 0.2 0.1
CA 0.4 0.5 0.8 0.6 0.5 0.7 0.4 0.3
Table 3
The matrix of subjective attribute weights
PC I QL FTZ CA
PC 1 3 7 7 9
I 1/3 1 3 3 7
QL 1/7 1/3 1 1 5
FTZ 1/7 1/3 1 1 5
CA 1/9 1/7 1/5 1/5 1
144 C. Kahraman et al. / Information Sciences 157 (2003) 135153
and the optimal alternative is dened as that achieving the highest degree of
membership in

DD [50].
The weights are used as exponents to express the importance of a goal. The
higher the importance of a goal the larger should be the exponent of its rep-
resenting fuzzy set, at least for normalized fuzzy sets and when using the min-
operator for the intersection of the fuzzy goals.
The solution procedure can be described as follows: Given the set
X = x
1
; . . . ; x
n
and the degrees of membership l
~
GG
j
(x
i
) of all x
i
in the fuzzy sets

GG
j
representing the goals [53].
1. Establish by pairwise comparison the relative importance, a
i
, of the goals
among themselves. Arrange the a
i
in a matrix M.
M =
a
1
a
1
a
1
a
2
. . .
a
1
a
n
a
2
a
1
a
2
a
2
. . .
. . .
a
n
a
1
a
n
a
2
. . .
a
n
a
n
_

_
_

_
(7)
2. Determine consistent weights w
j
for each goal by employing Saatys eigen-
vector method.
3. Weight the degrees of goal attainment, l
~
GG
j
(x
i
) exponentially by the respec-
tive w
j
. The resulting fuzzy sets are (

GG
j
(x
i
))
w
j
.
4. Determine the intersection of all (

GG
j
(x
i
))
w
j
:
~
DD = x
i
; min
j
(l
~
GG
j
(x
i
))
w
j
_ _

i
_
= 1; . . . ; n; j = 1; . . . ; m
_
(8)
5. Select the x
i
with largest degree of membership in

DD as the optimal alterna-
tive.
3.3.1. A numeric example of Yagers weighted goals method for the selection
among FLAs
For two FLAs, Table 4 gives evaluation ratings for each alternative re-
garding the attributes. Membership degrees are used for ratings. So, the sum
does not need to be equal to 1 for each attribute.
Using the same matrix of subjective attribute weights in Table 3 and Saatys
method, the following vector is again obtained [27]:
w
aw
= (0:537; 0:230; 0:100; 0:100; 0:033)
Exponential weighting of the attributes by their respective weight yields
P

CC
0:537
I;II
= (I; 0:826); (II; 0:611)

II
0:230
I;II
= (I; 0:872); (II; 0:906)
C. Kahraman et al. / Information Sciences 157 (2003) 135153 145
Q

LL
0:100
I;II
= (I; 0:978); (II; 0:912)
FT

ZZ
0:100
I;II
= (I; 0:912); (II; 0:965)
C

AA
0:033
I;II
= (I; 0:974); (II; 0:991)
The fuzzy set decision

DD, as the intersection of the sets above becomes

DD = (I; 0:826); (II; 0:611)


and the optimal alternative regarding subjective measures is alternative I with a
degree of membership in

DD of l
~
DD
(I) = 0:826.
3.4. Fuzzy AHP in group decisions
There are many fuzzy AHP methods proposed by various authors. These
methods are systematic approaches to the alternative selection and justication
problem by using the concepts of fuzzy set theory and hierarchical structure
analysis. Decision-makers usually nd that it is more condent to give interval
judgments than xed value judgments. This is because usually he/she is unable
to explicit about his/her preferences due to the fuzzy nature of the comparison
process.
The earliest work in fuzzy AHP appeared in van Laarhoven and Pedrycz
[48], which compared fuzzy ratios described by triangular membership func-
tions. Buckley [10] determines fuzzy priorities of comparison ratios whose
membership functions trapezoidal. Chang [16] introduces a new approach for
handling fuzzy AHP, with the use of triangular fuzzy numbers for pairwise
comparison scale of fuzzy AHP, and the use of the extent analysis method for
the synthetic extent values of the pairwise comparisons. Chan et al. [13] present
a technology selection algorithm to quantify both tangible and intangible
benets in fuzzy environment. They describe an application of the theory of
fuzzy sets to hierarchical structural analysis and economic evaluations. By
aggregating the hierarchy, the preferential weight of each alternative technol-
ogy is found, which is called fuzzy appropriate index. The fuzzy appropriate
Table 4
Evaluation ratings with membership degrees
Subjective attribute Alternatives
FLA
I
FLA
II
PC 0.7 0.4
I 0.55 0.65
QL 0.8 0.4
FTZ 0.4 0.7
CA 0.45 0.75
146 C. Kahraman et al. / Information Sciences 157 (2003) 135153
indices of dierent technologies are then ranked and preferential ranking or-
ders of technologies are found. From the economic evaluation perspective, a
fuzzy cash ow analysis is employed. Chan et al. [14] report an integrated
approach for the automatic design of FMS, which uses simulation and multi-
criteria decision-making techniques. The design process consists of the con-
struction and testing of alternative designs using simulation methods. The
selection of the most suitable design (based on AHP) is employed to analyze
the output from the FMS simulation models. Intelligent tools (such as expert
systems, fuzzy systems and neural networks) are developed for supporting the
FMS design process. Active X technique is used for the actual integration of
the FMS automatic design process and the intelligent decision support process.
Leung and Cao [35] propose a fuzzy consistency denition with consideration
of a tolerance deviation. Essentially, the fuzzy ratios of relative importance,
allowing certain tolerance deviation, are formulated as constraints on the
membership values of the local priorities. The fuzzy local and global weights
are determined via the extension principle. The alternatives are ranked on the
basis of the global weights by application of maximum-minimum set ranking
method. Kuo et al. [32] develop a decision support system for locating a new
convenience store. The rst component of the proposed system is the hierar-
chical structure development for fuzzy analytic process.
In the following, rst the outlines of the extent analysis method on fuzzy
AHP are given and then the method is applied to a facility location selection
problem.
Let X = x
1
; x
2
; . . . ; x
n
be an object set, and U = u
1
; u
2
; . . . ; u
m
be a
goal set. According to the method of Changs [15] extent analysis, each object
is taken and extent analysis for each goal is performed respectively. Therefore,
m extent analysis values for each object can be obtained, with the following
signs:
M
1
g
i
; M
2
g
i
; . . . ; M
m
g
i
; i = 1; 2; . . . ; n (9)
where all the M
j
g
i
(j = 1; 2; . . . ; m) are triangular fuzzy numbers.
The value of fuzzy synthetic extent with respect to the ith object is dened as
S
i
=

m
j=1
M
j
g
i

n
i=1

m
j=1
M
j
g
i
_ _
1
(10)
The degree of possibility of M
1
PM
2
is dened as
V (M
1
PM
2
) = sup
x Py
[min(l
M
1
(x); l
M
2
(y)[ (11)
When a pair (x; y) exists such that x Py and l
M
1
(x) = l
M
2
(y), then we have
V (M
1
PM
2
) = 1. Since M
1
and M
2
are convex fuzzy numbers we have that
V (M
1
PM
2
) = 1 iff m
1
Pm
2
(12)
C. Kahraman et al. / Information Sciences 157 (2003) 135153 147
V (M
1
PM
2
) = hgt(M
1
M
2
) = l
M
1
(d) (13)
where d is the ordinate of the highest intersection point D between l
M
1
and l
M
2
(see Fig. 2).
When M
1
= (l
1
; m
1
; u
1
) and M
2
= (l
2
; m
2
; u
2
), the ordinate of D is given by
Eq. (14).
V (M
2
PM
1
) = hgt(M
1
M
2
) =
l
1
u
2
(m
2
u
2
) (m
1
l
1
)
(14)
To compare M
1
and M
2
, we need both the values of V (M
1
PM
2
) and
V (M
2
PM
1
) The intersection between M
1
and M
2
is shown in Fig. 2.
The degree possibility for a convex fuzzy number to be greater than k convex
fuzzy numbers M
i
(i = 1; 2; . . . ; k) can be dened by
V (M PM
1
; M
2
; . . . ; M
k
)
= V [(M PM
1
) and M PM
2
) and . . . and (M PM
k
)[
= min V (M PM
i
); i = 1; 2; 3; . . . ; k (15)
Assume that
d
/
(A
i
) = min V (S
i
PS
k
) (16)
For k = 1; 2; . . . ; n; k ,= i. Then the weight vector is given by
W
/
= (d
/
(A
1
); d
/
(A
/
2
); . . . ; d
/
(A
n
))
T
(17)
where A
i
(i = 1; 2; . . . ; n) are n elements.
Via normalization, the normalized weight vectors are
W = (d(A
1
); d(A
2
); . . . ; d(A
n
))
T
(18)
where W is a nonfuzzy number.
M
2
M
1
l
2
m
1
u
1
u
2
d l
1
m
2
V ( M
2
M
1
)
1

Fig. 2. The intersection between M


1
and M
2
.
148 C. Kahraman et al. / Information Sciences 157 (2003) 135153
3.4.1. A numeric example to facility location selection using fuzzy AHP in group
decisions
A Turkish Motors Company, NEKYEK, considers investing in a new fac-
tory. The goal is to select the best location among the alternatives. The deci-
sion-making group consists of three members. The criteria taken into account
are environmental regulation (ER), host community (HC), competitive ad-
vantage (CA), and political risk (PR). The alternatives locations are Istanbul,
Ankara, and Izmir. The hierarchy is shown in Fig. 3. The fuzzy evaluation
matrix relevant to the goal is given in Table 5. The fuzzy scores are the com-
promise values of three members.
The weight vector from Table 5 is calculated as W = (0:19; 0:04; 0:77; 0:00).
The decision-making group now compares the alternatives with respect to the
decision criteria. Tables 69 show the evaluation matrixes.
The weight vector from Table 6 is calculated as W
ER
= (0:00; 0:43; 0:57).
The weight vector from Table 7 is calculated as W
HC
= (0:10; 0:56; 0:34).
The weight vector from Table 8 is calculated as W
CA
= (0:16; 0:39; 0:45).
The weight vector from Table 9 is calculated as W
PR
= (0:10; 0:56; 0:34).
Finally, adding the weights for location alternatives multiplied by the
weights of the corresponding criteria, a nal score is obtained for each location
alternative. Table 10 shows the nal scores for the location alternatives.
Izmir is selected for the new motor factory.
The best
facility

ER HC CA PR
Istanbul Ankara Izmir
Fig. 3. The hierarchy of the problem.
Table 5
The fuzzy evaluation matrix relevant to the goal
ER HC CA PR
ER (1, 1, 1) (3/2, 2, 5/2) (2/7, 1/3, 2/5) (5/2, 3, 7/2)
HC (2/5, 1/2, 2/3) (1, 1, 1) (2/7, 1/3, 2/5) (7/2, 4, 9/2)
CA (5/2, 3, 7/2) (5/2, 3, 7/2) (1, 1, 1) (5/2, 3, 7/2)
PR (2/7, 1/3, 2/5) (2/9, 1/4, 2/7) (2/7, 1/3, 2/5) (1, 1, 1)
C. Kahraman et al. / Information Sciences 157 (2003) 135153 149
4. Conclusions
Decisions are made today in increasingly complex environments. In more
and more cases the use of experts in various elds is necessary, dierent value
systems are to be taken into account, etc. In many of such decision-making
settings the theory of group decision-making (social choice) can be of use.
Fuzzy group decision-making can overcome this diculty. In general, many
concepts, tool and techniques of articial intelligence, in particular in the eld
of knowledge representation and reasoning, can be used to improve human
Table 6
Evaluation of alternatives with respect to ER
Istanbul Ankara Izmir
Istanbul (1, 1, 1) (2/5, 1/2, 2/3) (2/5, 1/2, 2/3)
Ankara (3/2, 2, 5/2) (1, 1, 1) (1/2, 2/3, 1)
Izmir (3/2, 2, 5/2) (1, 3/2, 2) (1, 1, 1)
Table 7
Evaluation of alternatives with respect to HC
Istanbul Ankara Izmir
Istanbul (1, 1, 1) (2/5, 1/2, 2/3) (1/2, 2/3, 1)
Ankara (3/2, 2, 5/2) (1, 1, 1) (1, 3/2, 2)
Izmir (3/2, 2, 5/2) (1, 3/2, 2) (1, 1, 1)
Table 8
Evaluation of alternatives with respect to CA
Istanbul Ankara Izmir
Istanbul (1, 1, 1) (1/2, 2/3, 1) (2/5, 1/2, 2/3)
Ankara (1, 3/2, 2) (1, 1, 1) (1/2, 1, 3/2)
Izmir (3/2, 2, 5/2) (2/3, 1, 2) (1, 1, 1)
Table 9
Evaluation of alternatives with respect to PR
Istanbul Ankara Izmir
Istanbul (1, 1, 1) (2/5, 1/2, 2/3) (1/2, 2/3, 1)
Ankara (3/2, 2, 5/2) (1, 1, 1) (1, 3/2, 2)
Izmir (1, 3/2, 2) (1/2, 2/3, 1) (1, 1, 1)
Table 10
The nal scores
Istanbul Ankara Izmir
Final scores 0.13 0.40 0.47
150 C. Kahraman et al. / Information Sciences 157 (2003) 135153
consistency and implementability of numerous models and tools in broadly
perceived decision-making and operations research.
Although four approaches have the same objective of selecting the best
facility location alternative, they come from dierent theoretic backgrounds
and relate dierently to the discipline of multi-attribute group decision-making.
Because the data needed by each approach is dierent from the others, we do
not necessarily expect to have the same result for the same facility location
problem.
Blins fuzzy relations use the relative popularity of an alternative over an-
other and then the nal non-fuzzy group preference. Fuzzy synthetic evaluation
uses crisp AHP to determine the relative importance of the criteria and then
calculates a grade for a specic alternative using the composition operation.
AHP employs a ratio scale to elicit pairwise comparisons of alternatives on each
criterion and an additive aggregation to global weights. Normalization of the
decision matrix is necessary to handle dierent types of attributes. Yagers
weighted goals method also uses crisp AHP to determine weights for each goal
and then weights the degrees of goal attainment exponentially. Fuzzy AHP
decides on the relative importance of each pair of factors in the same hierarchy.
It constructs the fuzzy evaluation matrix by using TFNs. Using the values of
extent analysis, it computes the value of fuzzy synthetic extent for each object.
From the point of view of the data needed by the approaches, while Blins
fuzzy relations are only based on social preference relation, fuzzy synthetic
evaluation is based on fuzzy relations between alternatives and evaluation
criteria and subjective attribute weights. Yagers weighted goals method is only
based on evaluation ratings and subjective attribute weights. Fuzzy AHP is
only based on fuzzy subjective attribute weights. From the point of view of
computational complexity, while Blins fuzzy relations are the least complex
among all, fuzzy synthetic evaluation and Yagers weighted goals method have
almost equal complexity. Fuzzy AHP has many computational steps and is the
most complex one among all.
References
[1] W. Applebaum, The analog method for estimating potential store sales, in: C. Kornblau (Ed.),
Guide to Store Location Research, Addison-Wesley, Reading, MA, 1968.
[2] M.A. Badri, Combining the analytic hierarchy process and goal programming for global
facility locationallocation problem, International Journal of Production Economics 62 (3)
(1999) 237248.
[3] M.J. Bender, S.P. Simonovic, A fuzzy compromise approach to water resource systems
planning under uncertainty, Fuzzy Sets and Systems 115 (2000) 3544.
[4] C.O. Benjamin, S.-C. Chi, T. Gaber, C.A. Riordan, Comparing BP and ART II neural network
classiers for facility location, Computers & Industrial Engineering 28 (1) (1995) 4350.
[5] O. Berman, Z. Drezner, G.O. Wesolowsky, Location of facilities on a network with groups of
demand points, IIE Transactions 33 (8) (2001) 637648.
C. Kahraman et al. / Information Sciences 157 (2003) 135153 151
[6] J. Bezdek, B. Spillman, R. Spillman, A fuzzy relation for group decision theory, Fuzzy Sets and
Systems 1 (1978) 255268.
[7] J. Bezdek, B. Spillman, R. Spillman, Fuzzy relation spaces for group decision theory: an
application, Fuzzy Sets and Systems 2 (1979) 514.
[8] J.M. Blin, Fuzzy relations in group decision theory, Journal of Cybernetics 4 (1974) 1722.
[9] J.M. Blin, A.P. Whinston, Fuzzy sets and social choice, Journal of Cybernetics 3 (1973) 2833.
[10] J.J. Buckley, Fuzzy hierarchical analysis, Fuzzy Sets and Systems 17 (1985) 233247.
[11] S.E. Butt, T.M. Cavalier, Ecient algorithm for facility location in the presence of forbidden
regions, European Journal of Operational Research 90 (1) (1996) 5670.
[12] C. Canel, B.M. Khumawala, J. Law, A. Loh, Algorithm for the capacitated, multi-commodity
multi-period facility location problem, Computers and Operations Research 28 (5) (2001) 411
427.
[13] F.T.S. Chan, M.H. Chan, N.K.H. Tang, Evaluation methodologies for technology selection,
Journal of Materials Processing Technology 107 (2000) 330337.
[14] F.T.S. Chan, B. Jiang, N.K.H. Tang, The development of intelligent decision support tools to
aid the design of exible manufacturing systems, International Journal of Production
Economics 65 (1) (2000) 7384.
[15] D.-Y. Chang, in: Extent Analysis and Synthetic Decision, Optimization Techniques and
Applications, vol. 1, World Scientic, Singapore, 1992, p. 352.
[16] D.-Y. Chang, Applications of the extent analysis method on fuzzy AHP, European Journal of
Operational Research 95 (1996) 649655.
[17] R.B. Chase, N. Aquilano, Production and Operations Management, Richard D. Irwin, Inc,
Chicago, 1995.
[18] S.-M. Chen, Fuzzy group decision-making for evaluating the rate of aggregative risk in
software development, Fuzzy Sets and Systems 118 (2001) 7588.
[19] C.T. Chen, A fuzzy approach to select the location of the distribution center, Fuzzy Sets and
Systems 118 (2001) 6573.
[20] W.J. Comley, Location of ambivalent facilities: use of a quadratic zero-one programming
algorithm, Applied Mathematical Modeling 19 (1) (1995) 2629.
[21] F. Herrera, E. Herrera-Viedma, J.L. Verdegay, A model of consensus in group decision-
making under linguistic assessments, Fuzzy Sets and Systems 79 (1996) 7387.
[22] F. Herrera, E. Herrera-Viedma, J.L. Verdegay, Direct approach processes in group decision-
making using linguistic OWA operators, Fuzzy Sets and Systems 79 (1996) 175190.
[23] F. Herrera, E. Herrera-Viedma, J.L. Verdegay, Linguistic measures based on fuzzy coincidence
for reaching consensus in group decision-making, International Journal of Approximate
Reasoning 16 (1997) 309334.
[24] H.-M. Hsu, C.-T. Chen, Aggregation of fuzzy opinions under group decision-making, Fuzzy
Sets and Systems 79 (1996) 279285.
[25] C. Jungthirapanich, An intelligent decision support system for facility location, Unpublished
Dissertation, Department of Engineering Management, University of Missouri-Rolla, Rolla,
MO 65401, 1992.
[26] J. Kacprzyk, M. Fedrizzi, H. Nurmi, Group decision-making with fuzzy majorities represented
by linguistic quantiers, in: D. Dubois, H. Prade, R.R. Yager (Eds.), Readings in Fuzzy Sets
for Intelligent Systems, Morgan Kaufmann Publishers, Inc, 1993.
[27] C. Kahraman, Z. Ulukan, E. Tolga, A fuzzy weighted evaluation method using objective and
subjective measures, in: Proceedings of International ICSC Symposium on Engineering of
Intelligent Systems (EIS98), vol. 1, University of La Laguna Tenerife, Spain, 1998, pp. 5763.
[28] C. Kahraman, E. Tolga, Group decision-making for the selection among CIM systems using
fuzzy concept, in: Proceedings of the IASTED International Conference on Articial
Intelligence and Soft Computing (ASC98), Cancun, Mexico, 1998, pp. 465468.
152 C. Kahraman et al. / Information Sciences 157 (2003) 135153
[29] W. Karwowski, A. Mital, Potential applications of fuzzy sets in industrial safety engineering,
Fuzzy Sets and Systems 19 (1986) 105120.
[30] G.J. Klir, A.T. Folger, Fuzzy Sets, Uncertainty, and Information, Prentice-Hall, Inc, 1988.
[31] A. Klose, LP-based heuristic for two-stage capacitated facility location problem, Journal of the
Operational Research Society 50 (2) (1999) 157166.
[32] R.J. Kuo, S.C. Chi, S.S. Kao, A decision support system for selecting convenience store
location through integration of fuzzy AHP and articial neural network, Computers in
Industry, 2002, in press.
[33] R.J. Kuo, S.C. Chi, S.S. Kao, A decision support system for locating convenience store
through fuzzy AHP, Computers & Industrial Engineering 37 (1999) 323326.
[34] H.-M. Lee, Group decision-making using fuzzy sets theory for evaluating the rate of
aggregative risk in software development, Fuzzy Sets and Systems 80 (1996) 261271.
[35] L.C. Leung, D. Cao, On consistency and ranking of alternatives in fuzzy AHP, European
Journal of Operational Research 124 (2000) 102113.
[36] R.-J. Li, Fuzzy method in group decision-making, Computers and Mathematics with
Applications 38 (1999) 91101.
[37] S. Melkote, M.S. Daskin, Capacitated facility location/network design problems, European
Journal of Operational Research 129 (3) (2001) 481495.
[38] F.J. Montero, J. Tejada, Some problems in the denition of fuzzy preference relations, Fuzzy
Sets and Systems 6 (1986) 249259.
[39] S. Nanthavanij, P. Yenradee, Predicting the optimum number, location, and signal sound level
of auditory warning devices for manufacturing facilities, International Journal of Industrial
Ergonomics 24 (6) (1999) 569578.
[40] S. Nobuaki, U. Akira, D. Atsushi, O. Akira, S. Seiji, H. Satoshi, Commercial facility location
model using multiple regression analysis, Computers Environment and Urban Systems 22 (3)
(1998) 219240.
[41] H. Nurmi, Approaches to collective decision-making with fuzzy preference relations, Fuzzy
Sets and Systems 6 (1981) 249259.
[42] T.J. Ross, Fuzzy Logic with Engineering Applications, McGraw-Hill, New York, 1995.
[43] T.L. Saaty, The Analytic Hierarchy Process, McGraw-Hill, New York, 1980.
[44] T.L. Saaty, Group decision-making and the AHP, in: B.L. Golden, E.A. Wasil, P.T. Harker
(Eds.), The Analytic Hierarchy Process: Applications and Studies, Springer, New York, 1989,
pp. 5967.
[45] T. Tanino, Fuzzy preference orderings in group decision-making, Fuzzy Sets and Systems 12
(1984) 117131.
[46] T. Tanino, Fuzzy preference relations in group decision-making, in: Kacprzyk and Roubens,
1988, pp. 5471.
[47] G.-H. Tzeng, Y.-W. Chen, Optimal location of airport re stations: a fuzzy multi-objective
programming and revised genetic algorithm approach, Transportation Planning and Tech-
nology 23 (1) (1999) 3755.
[48] P.J.M. van Laarhoven, W. Pedrycz, A fuzzy extension of Saatys priority theory, Fuzzy Sets
and Systems 11 (1983) 229241.
[49] J. Wang, B. Malakooti, A feedforward neural network for multiple criteria decision-making,
Computers in Operations Research 19 (2) (1992) 151167.
[50] R.R. Yager, Fuzzy decision-making including unequal objectives, Fuzzy Sets and Systems 1
(1978) 8795.
[51] R.R. Yager, Fuzzy Sets and Possibility Theory, Pergamon Press, Oxford, 1982.
[52] L.A. Zadeh, A theory of commonsense knowledge, Memo UCB/ERL M83/26, University of
California, Berkeley, 1983.
[53] H.-J. Zimmermann, Fuzzy Sets and its Applications, Kluwer Publishing, 1994.
C. Kahraman et al. / Information Sciences 157 (2003) 135153 153