Expert Systems with Applications 38 (2011) 10803–10811

Contents lists available at ScienceDirect

Expert Systems with Applications

journal homepage: www.elsevier.com/locate/eswa

An integrated fuzzy TOPSIS and MCGP approach to supplier selection in supply

chain management
Chin-Nung Liao a,b,⇑, Hsing-Pei Kao b,1
a
Department of Business Administration, China University of Science and Technology, No. 245, Sec. 3, Academia Rd., Nankang, Taipei 115, Taiwan, ROC
b
Graduate Institute of Industrial Management, National Central University, No. 300, Jhongda Rd., Jhongli City, Taoyuan County 320, Taiwan, ROC

a r t i c l e i n f o a b s t r a c t

Keywords:
Fuzzy numbers
Supply chain management
Supplier selection
Multiple criteria decision making (MCDM)
Techniques for order preference by
similarity to ideal solution (TOPSIS)
Multi-choice goal programming (MCGP)
Supplier selection is an important issue in supply chain management. In recent years, determining the
best supplier in the supply chain has become a key strategic consideration. However, these decisions usu-
ally involve several objectives or criteria, and it is often necessary to compromise among possibly con-
flicting factors. Thus, the multiple criteria decision making (MCDM) becomes a useful approach to
solve this kind of problem. Considering both tangible and intangible criteria, this study proposes inte-
grated fuzzy techniques for order preference by similarity to ideal solution (TOPSIS) and multi-choice
goal programming (MCGP) approach to solve the supplier selection problem. The advantage of this
method is that it allows decision makers to set multiple aspiration levels for supplier selection problems.
The integrated model is illustrated by an example in a watch firm.
Ó 2011 Published by Elsevier Ltd.

1. Introduction can essentially be considered a multiple criteria decision making

Supplier selection is an important issue in supply chain man-
agement. Typically, manufacturer spends more than 60% of its total
sales on purchased items, such as raw materials, parts, and compo-
nents (Krajewsld & Ritzman, 1996). In addition, manufacturer pur-
chases of goods and services constitute up to 70% of product cost
(Ghodsypour & O’Brien, 1998). Therefore, the selection of suppliers
is an area of tremendous importance and should be considered a
strategic issue in the effective management of a supply chain. Sup-
plier selection and its related tasks are positioned at the front end
in the supply chain process (see Fig. 1).
During the 1990s, many manufacturers sought to develop stra-
tegic alliances with suppliers in order to upgrade their manage-
ment preference and competitiveness (Kumar, Vrat, & Shankar,
2006; Shin, Collier, & Wilson, 2000). While coordination between
a manufacturer and its suppliers is typically an important and dif-
ficult link in the channel of distribution, many methods have been
adopted for supplier selection under rather simplistic perceptions
of the decision making process (Chen, Lin, & Huang, 2006). How-
ever, supplier evaluation and selection are complicated by the need
for decision makers (DMs) to consider various criteria.
The selection process mainly involves the evaluation of differ-
ent criteria and various supplier attributes. This selection process
⇑ Corresponding author. Tel.: +886 2 27821862x214; fax: +886 2 27864984.
E-mail addresses: lliao@cc.cust.edu.tw (C.-N. Liao), hpkao@mgt.ncu.edu.tw
(H.-P. Kao).
1
Tel.: +886 3 4227151x66154; fax: +886 3 4258197.
(MCDM) problem, which is affected by different tangible and
intangible criteria (Pi & Low, 2005). Since 1966, many criteria have
been employed to evaluate and select supplier. Dickson (1966)
identified 23 different criteria for supplier selection, based on
which Weber, Current, and Benton (1991) suggested a number of
selection criteria to measure supplier performance, such as price,
delivery, quality, productive capability, location, technical capabil-
ity, management organization, reputation, industry position, finan-
cial stability, performance history, and maintainability. Evans
(1980) proposed that price, quality and delivery are key criteria
for supplier evaluation in the industrial market. Shipley (1985)
suggested that supplier selection involve three criteria, namely,
quality, price and delivery lead time. Ellram (1990) suggested that
in the supplier selection process, firms must to consider whether
product quality, offering price, delivery time, and total service
quality meet organizational demand. Tam and Tummala (2001)
proposed an analytic hierarchy process (AHP) based model and
adopted quality, cost, problem-solving capabilities, expertise,
delivery lead time, response to customer requests, experience,
and reputation in selecting telecommunications systems. Pi and
Low (2005) suggested a method for supplier evaluation and selec-
tion based on quality, on-time delivery, price, and service quality.
Recently, the supplier selection process has received consider-
able attention in the marketing management literature. Chen
et al. (2006) adopted a fuzzy decision making approach to address
the supplier selection problem in the supply chain system. Five
benefit criteria were considered, including the profitability of sup-
plier, relationship closeness, technological capability, conformance

0957-4174/$ - see front matter Ó 2011 Published by Elsevier Ltd.

doi:10.1016/j.eswa.2011.02.031
10804 C.-N. Liao, H.-P. Kao / Expert Systems with Applications 38 (2011) 10803–10811

The phase of supplier selection

Manufacturer / Company Customer

Suppliers Retailers

Customer
Suppliers Procurement Produce Marketing Retailers

Suppliers Retailers
Customer

Fig. 1. The flow process of supply chain management.

quality, and conflict resolution. Lin and Chang (2008) claimed that
communication, reputation, industry position, relationship close-
ness, customer responsiveness, and conflict-solving capabilities
are important criteria in vendor selection. In addition, the role of
organizational size in the supplier selection process has been ad-
dressed by Wang, Cheng, and Cheng (2009). Table 1 summarizes
the criteria that have appeared in literature since 1966; most of
the articles referenced above suggest that quality, price, and deliv-
ery performance are the most important supplier selection criteria.
Over the years, a number of techniques have been proposed to
solve the supplier selection problem. The long list of approaches
includes linear programming (LP), mathematical programming
models, multiple-objective programming, statistical and probabi-
listic methods, data envelopment analysis (DEA), cost-based meth-
ods (CBM), case-based reasoning (CBR), neural networks (NN),
AHP, analytic network process (ANP), fuzzy set theory, and tech-
niques for order preference by similarity to ideal solution (TOPSIS).
Recently, the integration of different methodologies to supplier
selection process has received considerable attention in the supply
chain management literature. Faez, Ghodsypour, and O’brien
(2009) presented an integrated fuzzy case-based reasoning and
mathematical programming method. Önüt, Kara, and Isik (2009)
developed a supplier evaluation approach based on the ANP and
TOPSIS methods to help a telecommunication company in vendor
selection. Ha and Krishnan (2008) developed a hybrid model that
including AHP, DEA and NN approaches to the supplier selection
problem. Most recently, Kokangul and Susuz (2009) integrated
AHP and mathematical programming to consider both non-linear
integer and multiple-objective programming under certain con-
straints to determine the best suppliers. The integrated model uses
source data provided by a manufacturing firm to address a real-
world supplier selection problem.

Table 1
Supplier selection criteria literature review.

Selection criteria 1 2 3 4 5 6 7 8 9 10
p p p p p p
Price (cost)
p p p p p p p p
Product quality
p p p p p p p
On-time delivery
p
Warranty and claims
p p
After sales service
p
Technical support/expertise
p
Attitude
p p
Total service quality
p
Training aids
p p p
Performance history
p p p p
Financial stability
p p
Location
p
Labor relations
p p
Relationship closeness
p p
Management and organization
p p p
Conflict/problem solving capability
p p
Communication system
Response to customer request
p p p
Technical capability
p p
Production capability
p
Packaging capability
p
Operational controls
p
Amount of past business
p p p p p
Reputation and position in industry
p
Reciprocal arrangements
p
Impression
p
Business attempt
p p
Maintainability
p
Size

1, Dickson (1966); 2, Evans (1980); 3, Shipley (1985); 4, Ellram (1990); 5, Weber et al. (1991); 6, Tam and Tummala (2001); 7, Pi and Low (2005); 8, Chen et al. (2006); 9, Lin
and Chang (2008); 10, Wang et al. (2009).
 C.-N. Liao, H.-P. Kao / Expert Systems with Applications 38 (2011) 10803–10811 10805

In real life, the modeling of many situations may not be suffi-

cient or exact, as the available data are inexact, vague, imprecise
and uncertain by nature (Sarami, Mousavi, & Sanayei, 2009).
Moreover, the decision making processes that take place in such
situations are also based on uncertain and ill-defined information.
In the real practice of supplier selection, firms usually confronts
with a high degree of uncertainties and fuzziness. Fuzzy set theory
is considered the most effective methods in managing vagueness
and uncertainty problems. The concept of fuzzy sets was intro-
duced by Zadeh (1965) to mathematically represent data and
information possessing non-statistical uncertainties and to provide
formalized tools for dealing with imprecision intrinsic to many
problems (Kahraman, Cevik, Ates, & Güfer, 2007). In order to model
such situations, fuzzy set theory was introduced to express the lin- e.
Fig. 2. Trapezoidal fuzzy number n
guistic terms of decision marking processes.
In addition, TOPSIS is a classical MCDM method, and as such it
may provide the basis for developing supplier selection models Definition 2.2. A matrix D ~ is called a fuzzy matrix, if it contains at
that can effectively handle uncertainty properties. This approach ~
least an entry in D is a fuzzy number (Buckley, 1985).
is based on the idea that a chosen alternative should be the short-
est distance from the positive-ideal solution and the farthest dis-
Definition 2.3. Let m ~ ¼ ðn1 ; n2 ; n3 ; n4 Þ be
~ ¼ ðm1 ; m2 ; m3 ; m4 Þ and n
tance from the negative-ideal solution. Chen et al. (2006) applied
two triangular fuzzy numbers. Then the distance between them
linguistic value to measure the ratings and weights of supplier
can be calculated using the vertex method (see Chen, 2000; Chen
selection criteria and then used a MCDM model based on fuzzy
et al., 2006):
set theory to analyze a supply chain management case. However,
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
their model is only suitable for single-sourcing problems. In the
~ n
dðm; ~Þ ¼ ð1=4Þ½ðm1  n1 Þ2 þ ðm2  n2 Þ2 þ ðm3  n3 Þ2 þ ðm4 þ n4 Þ2 :
single sourcing scenario, one supplier presumably will satisfy the
buyer’s needs; such that the decision must be made on which sup- ð2Þ
plier is the best. In the multi-sourcing case, no supplier can satisfy
In this study, the importance weights of various criteria and the
all the buyer’s requirements, and so more than one supplier can be
ratings of qualitative criteria are considered linguistic variables.
selected (Ghodsypour & O’Brien, 1998).
Because linguistic assessments approximate the subjective judg-
In this study, an integrated fuzzy TOPSIS and MCGP model is
ment of DMs, we consider linear trapezoidal membership functions
developed to solve multi-sourcing supplier selection problems. First,
adequate for capturing the vagueness of these linguistic assess-
linguistic values expressed in trapezoidal fuzzy numbers are applied
ments. These linguistic variables can be expressed in positive trap-
to assess weights and ratings of supplier selection criteria. Second, a
ezoidal fuzzy numbers, as shown in Figs. 3 and 4. In the proposed
hierarchy multi-model based on fuzzy set theory is expressed and
integrated model, the DMs use the linguistic variables to evaluate
fuzzy positive and negative-ideal solutions are used to find each
the importance of criteria and the ratings of alternative suppliers
supplier’s closeness coefficient. Finally, a MCGP model based on
with respect to each selection criterion.
the tangible constrains regarding the buyer and its suppliers is
For example, referring to Fig. 3, ‘‘Medium’’ can be represented
constructed and solved to assign order qualities to each supplier.
as (0.4, 0.5, 0.6, 0.7); it has the following membership function:
The paper is organized as follows. The next section introduces 8
the basic definitions and notations of fuzzy numbers and linguistic >
> 0; x < 0:4;
>
>
variables. Section 3 presents both the GP and MCGP approaches. >
< x  0:4=0:5  0:4;
> 0:4 6 x 6 0:5;
Section 4 presents the analytical procedure of the proposed inte- uMedium ðxÞ ¼ 1; 0:5 6 x 6 0:6; ð3Þ
grated approach. In Section 5, the proposed method is illustrated >
>
>
> x  0:7=0:6  0:7; 0:6 6 x 6 0:7;
using a watch company as an example. The finally section presents >
>
:
conclusions and suggestions for future research. 0; x > 0:7:

2. Basic definitions and notation

Some basic concepts of fuzzy numbers and linguistic variables y

are now defined. Low Medium High

Very Medium Medium Very

Definition 2.1. A positive trapezoidal fuzzy number n ~ can be Low High
Low High
defined as (n1, n2, n3, n4) as shown in Fig. 2. The membership 1
function un~ ðxÞ is defined as follows:

8
>
> 0; x < n1 ;
>
>
>
< x  n1 =n2  n1 ;
> n1 6 x 6 n2 ;
un~ ðxÞ ¼ 1; n2 6 x 6 n3 ; ð1Þ
>
>
>
> x  n4 =n3  n4 ; n3 6 x 6 n4 ;
>
>
:
0; x > n4 :
x
For a trapezoidal fuzzy number n ~ ¼ ðn1 ; n2 ; n3 ; n4 Þ, when n2 = n2, the
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
number is called a triangular fuzzy number. A crisp number k can be
expressed as (k, k, k, k). Fig. 3. Linguistic variables for importance weight of each criterion.
10806 C.-N. Liao, H.-P. Kao / Expert Systems with Applications 38 (2011) 10803–10811

As stated in Definition 2.2, a supplier selection problem can be

y expressed in a matrix as follows (Chen et al., 2006):
Poor Fair Good 2 3
~x11 ~x12  ~x1n
Very Very
Poor
Medium Medium 6 ~x ~x22    ~x2n 7
Poor Good Good 6 21 7
~¼6
.. 7
X ~ ¼ ½w
W ~ 1; w
~ 2; . . . ; w
~ n ; ð8Þ
1 6 .. .. ... 7;
4 . . . 5
~xm1 ~xm2    ~xmn

where ~xij ¼ ðaij ; bij ; cij ; dij Þ and w

~ j ¼ ðwj1 ; wj2 ; wj3 ; wj4 Þ; i ¼ 1; 2; 3; . . .
m, j ¼ 1; 2; 3; . . . ; n.
According to the briefly-summarized discussion of fuzzy set
theory, the normalized fuzzy decision matrix can be represented
as:

x ~ ¼ ½~r ij ;
R i ¼ 1; 2; 3; . . . m; j ¼ 1; 2; 3; . . . ; n; ð9Þ
1 2 3 4 5 6 7 8 9 10
where the ~rij is the normalized value of ~xij ¼ ðaij ; bij ; cij ; dij Þ, which be
Fig. 4. Linguistic variables for ratings. calculated as follows:
In addition, the linguistic variable ‘‘Good’’ can be represented as If the jth criterion is a benefit, then:
(7, 8, 8, 9) in Fig. 4; its membership function is: ~rij ¼ ðaij =dj ; bij =dj ; cij =dj ; dij =dj Þ; ð10Þ
8
< 0;
> x < 7;
where

dj
¼ max dij .
uGood ðxÞ ¼ x  7=8  7; 7 6 x 6 8; ð4Þ If the jth criterion is a cost, then:
>
:
1; 8 6 x 6 9:
~rij ¼ ðaj =dij ; aj =cij ; aj =bij ; aj =aij Þ; ð11Þ
Assume that a decision group has K DMs and the fuzzy ratings
of all decision maker preferences are trapezoidal fuzzy member, where aj ¼ min aij .
~ k ¼ ðak ; bk ; ck ; dk Þ, k ¼ 1; 2; . . . ; K. Then the aggregated fuzzy rating
R A weighted normalized fuzzy-decision matrix can be con-
can be defined as: structed according to the normalized fuzzy-decision matrix as
follows:
~ ¼ ða; b; c; dÞ;
R k ¼ 1; 2; . . . ; K ð5Þ
~ ¼ ½v~ ij  ;
V ð12Þ
mn
where
where v~ ij ¼ x~ij  w
~ j ; i ¼ 1; 2; 3; . . . m; j ¼ 1; 2; 3; . . . ; n:
X
K
After constructing a weighted normalized fuzzy-decision ma-
aij ¼ minfak g; b ¼ 1=K bK ;
k
K¼1
trix, the fuzzy positive-ideal solution (FPIS), S , and the fuzzy neg-
ative-ideal solution (FNIS), S , can be calculated as follows:
X
K 
c ¼ 1=K cK ; d ¼ maxfdk g: S ¼ fðmax v~ ij jj 2 J Þ; ðmin v~ ij j 2 J 0 Þg ¼ ðv~ 1 ; v~ 2 ; . . . ; v~ n Þ;
k i i
K¼1
i ¼ 1; 2; . . . ; m; j ¼ 1; 2; 3; . . . ; n ð13Þ
Let the fuzzy rating and importance weight of the kth DM be
~
xijk ¼ ðaijk ; bijk ; cijk ; dijk Þ and w ~ jk ¼ ðw
~ jk1 ; w
~ jk2 ; w
~ jk3 ; w
~ jk4 Þ, where i ¼ and
1; 2; . . . ; m, j ¼ 1; 2; 3; . . . ; n, respectively. Therefore, the aggregated 
S ¼ fðmin v~ ij jj 2 J Þ; ðmax v~ ij j 2 J 0 Þg ¼ ðv~ 1 ; v~ 2 ; . . . ; v~ n Þ;
fuzzy ratings, ~ xij ; of alternatives with respect to each criterion can i i
be calculated as (Chen et al., 2006): i ¼ 1; 2; . . . ; m; j ¼ 1; 2; 3; . . . ; n; ð14Þ
~xij ¼ ðaij ; bij ; cij ; dij Þ; ð6Þ
where v ¼ maxfv ij4 g and v ¼ minfv ij1 g. In addition, J is associ-

j
~
j
where ated with benefit criteria, while J 0 is associated with cost criteria.
The distance of each alternative from S and S can be calcu-
X
K
aij ¼ minfaijk g; bij ¼ 1=K bijk ; lated as:
k
K¼1

X
n
di ¼ dðv~ ij; v~ þj Þ; i ¼ 1; 2; . . . ; m; ð15Þ
X
K
j¼1
cij ¼ 1=K cijk ; dij ¼ maxfdijk g;
k
K¼1

X
n
~ j , of each criterion can be calcu-
and the aggregated fuzzy weights, w di ¼ dðv~ ij; v~ j Þ; i ¼ 1; 2; . . . ; m; ð16Þ
j¼1
lated as (Chen et al., 2006):
~ j ¼ ðwj1 ; wj2 ; wj3 ; wj4 Þ;
w ð7Þ where dð; Þ represents the distance measurement between two
fuzzy numbers.
where
Finally, the closeness coefficients of each supplier according to
X
K distance from the fuzzy positive-ideal solution (FPIS), S and the
wj1 ¼ minfwjk1 g; wj2 ¼ 1=K wjk2 ; fuzzy negative-ideal solution (FNIS), S , can be calculated as:
k
K¼1
  
CC i ¼ di =ðdi þ di Þ; i ¼ 1; 2; . . . ; m; ð17Þ
X
K
wj3 ¼ 1=K wjk3 ; wj4 ¼ maxfwjk4 g: where CC i range belongs to the closed interval [0, 1] and
k
K¼1 i ¼ 1; 2; . . . ; m.
 C.-N. Liao, H.-P. Kao / Expert Systems with Applications 38 (2011) 10803–10811 10807

3. Multi-choice goal programming where the coefficients of decision variables denote product prices;
x1, x2, and x3 represents three products, and target values (e.g., 90,
Goal programming (GP) is an analytical multiple objectives 100, and 110) are three profit goals, respectively.
decision making approach designed to address decision-marking Using a MCGP method, this problem can be expressed as the
problems in which targets have been assigned to all attributes following program:
and where the decision makers are interested in minimizing the þ  þ  þ 
Min z ¼ d1 þd1 þd2 þd2 þd3 þd3
non-achievement of a particular goal (Liao, 2009). The model can
þ 
take into account many simultaneous objectives as a decision ma- s:t: 2:5x1 þ3x2 þ2x3 d1 þd1 ¼ 90b1 þ120ð1b1 Þ;
ker seeks the best solution from among a set of feasible solutions. þ 
3x1 þ2x2 þx3 d2 þd2 ¼ 80b2 þ100ð1b2 Þ;
GP was first introduced by Charnes and Cooper (1961), and has þ 
been further developed by Ignizio (1976), Zimmermann (1978),
3:5x1 þ3x2 þ5x3 d3 þd3 ¼ 75b3 b4 þ90b3 ð1b4 Þþ110ð1b3 Þb4 ;
Tamiz, Jones, and Romero (1998), Romero (2001), Chang (2007) x1 þx2 þx3 P 17;
and Liao (2009), and so on. x2 þx3 P 12;
Initially, the GP was expressed as follows: x2 P 5;
(GP model) þ 
di ; di P 0; i ¼ 1;2;3:
X
n
Min jfi ðXÞ  g i j ð18Þ þ 
i¼1 where b1 ; b2 ; b3 and b4 are binary variables, and di and di are the
s:t: X 2 F ðF is a feasible setÞ; positive and negative deviations of ith goal, respectively.
This problem can be solved using the existing optimization
where fi ðXÞ is the linear function of the ith goal, and g i is the aspi- software LINGO (Schrage, 2002) to obtain the optimal solutions
ration level of the ith goal.  16:42;
ðx1 ; x2 ; x3 ; b1 ; b2 ; b3 ; b4 Þ ¼ ð15:31;  1:23;
 1; 1; 0; 1Þ: From
The above minimization process can be accomplished using the these results, we see that goal G1 reaches the aspiration level 90
weighted GP (WGP) model as follows: at 90; goal G2 reaches the aspiration level 80 at 80; and goal G3
(WGP model) reaches the aspiration level 110 at 108.99.
X
n
þ 
Consider the following multiple-objective decision-making
Min ðai di þ bi di Þ ð19Þ problems, modified slightly from Example 1.
i¼1
 þ
s:t: f i ðXÞ þ di di ¼ g i ; i ¼ 1; 2; . . . ; n; Example 2
þ 
di ; di  0; i ¼ 1; 2; . . . ; n;
Goals : ðG1 Þ 2:5x1 þ 3x2 þ 2x3 ¼ 90;
X 2 F ðF is a feasible setÞ;
ðG2 Þ 3x1 þ 2x2 þ x3 ¼ 80;
where parameters ai and bi are the weights reflecting preferential ðG3 Þ 3:5x1 þ 3x2 þ 5x3 ¼ 90:
and normalizing purposes attached to positive and negative devia-
 Constraints : x1 þ x2 þ x3  17; x2 þ x3  12; x2  5:
tions of ith goal, respectively; di ¼ maxð0; g i  fi ðXÞÞ and
þ
di ¼ maxð0; fi ðXÞ  g i Þ are, respectively, under- and over-achieve- where the coefficients of decision variables and target values are
ments of the ith goal, fi ðXÞ and g i are defined as above in the GP defined as in Example 1.
model (Ignizio, 1976).
GP approaches have been applied to solve many real-world
problems. However, many multiple-choice aspiration levels may Again, this problem can be solved using optimization software
exist, such as ‘‘something more/higher is better’’ or ‘‘something LINGO (Schrage, 2002) to obtain the optimal solutions
 0Þ: From these results, we see that goal G1
ðx1 ; x2 ; x3 Þ ¼ ð15; 17:5;
less/lower is better’’ (Chang, 2007). These typical multiple choice
GP problems cannot be solved using a traditional GP approach. reaches the aspiration level 90 at 90; goal G2 reaches the aspiration
Chang (2007) presented a MCGP method to solve these types of level 80 at 80; and goal G3 has a positive value (+15) over aspira-
problems, and it can be expressed as follows: tion level 90.
(MCGP model) According to the above examples, Example 1 is better than
Example 2 for the decision maker, as the solution to Example 1 is
X
n
þ  perfectly balanced across the three goals. In other words, the more
Min ðai di þ bi di Þ ð20Þ
i¼1
the aspiration levels there are, the better are the solutions gener-
þ  ated by this proposed MCGP method. This configuration forces an
s:t: f i ðXÞ  di þ di ¼ g i1 or g i2 or g i3 or    or g im ;
þ 
optimized consensus by minimizing total deviation (Chang, 2007).
di ; di  0; i ¼ 1; 2; . . . ; n; In addition, to reduce the binary variables, the MCGP can be
X 2 F ðF is a feasible setÞ; reformulated as the following two alternative MCGP achievement
where g ij ði ¼ 1; 2; . . . ; n and j ¼ 1; 2; . . . ; mÞ is the jth aspiration le- functions (Chang, 2008).
vel of the ith goal, g ij1 g ij g ijþ1 . Other variables are defined as Type 1: ‘‘the more the better’’ case:
in WGP. X
n
þ 
Below we present two examples of MCGP problem that have Min ½wi ðdi þ di Þ þ aðeþi þ ei Þ ð21Þ
i¼1
certain goals and constraints, while they cannot be solved by tradi-
þ 
tional GP approaches. s:t: f i ðXÞ  di þ di ¼ yi ; i ¼ 1; 2; . . . ; n; ð22Þ
yi  eþi þ ei ¼ g i;max ; i ¼ 1; 2; . . . ; n ð23Þ
Example 1 g i;min 6 yi g i;max ;
þ 
Goals : ðG1 Þ 2:5x1 þ 3x2 þ 2x3 ¼ 90 or 120; di ; di ; eþi ; ei  0; i ¼ 1; 2; . . . ; n;
ðG2 Þ 3x1 þ 2x2 þ x3 ¼ 80 or 100; X 2 F ðF is a feasible set; X is unrestricted in signÞ;
ðG3 Þ 3:5x1 þ 3x2 þ 5x3 ¼ 75; 90 or 110: þ 
where di and di are the positive and negative deviations corre-
Constraints : x1 þ x2 þ x3  17; x2 þ x3  12; x2  5: sponding to the ith goal jfi ðXÞ  yi j in Eq. (20); eþ 
i and ei are the
10808 C.-N. Liao, H.-P. Kao / Expert Systems with Applications 38 (2011) 10803–10811

positive and negative deviations corresponding to jyi  g i;max j in Eq. Step 4. Construct weighted normalized fuzzy-decision matrix.
(21); and ai is the weight attached to the sum of the deviation of Step 5. Determine FPIS and FNIS.
jyi  g i;max j. Other variables are defined as in MCGP. Step 6. Calculate the distance of each supplier from FPIS and FNIS,
respectively.
Type 2: ‘‘the less the better’’ case:
Step 7. Calculate the closeness coefficient (CCi) of each supplier.
X
n
þ  Step 8. According to the closeness coefficients obtained from Step
Min ½wi ðdi þ di Þ þ aðeþi þ ei Þ ð24Þ 7 for each supplier, build the integrated model to find the
i¼1
þ  best suppliers and their optimum order quantities. In
s:t: f i ðXÞ  di þ di ¼ yi ; i ¼ 1; 2; . . . ; n; ð25Þ order to find the best order quantities, the total value cre-
yi  eþi þ ei ¼ g i;min ; i ¼ 1; 2; . . . ; n; ð26Þ ated from the procurement (TVP) should be maximized.
g i;min yi g i;max ;
þ  The final model (FTM model) which integrates fuzzy TOPSIS and
di ; di ; eþi ; ei  0; i ¼ 1; 2; . . . ; n;
MCGP can be shown as:
X 2 F ðF is a feasible set; X is unrestricted in signÞ; Achievement functions
þ 
where di and di are the positive and negative deviations corre- X
n
þ 
sponding to the ith goal jfi ðXÞ  yi j in Eq. (23); eþ  Min ðdi þ di Þ ð27Þ
i and ei are the po-
i¼1
sitive and negative deviations corresponding to jyi  g i;min j in Eq.
(24); and ai is the weight attached to the sum of the deviation of s:t:
jyi  g i;min j. Other variables are defined as in MCGP.
Goal and systems constraints
X
n
þ 
4. The proposed method CC i  X i  d1 þ d1 P g 1;min ðTVP goalÞ ð28Þ
i¼1

The proposed method not only considers DM’s preference and

experience for supplier selection criteria, but also includes various X
n
þ 
pricei X i  d2 þ d2 P g 2;min or 6 g 2;max ðBudget constraintÞ
tangible constraints, for example, the buyer’s budget, suppliers’ i¼1
capacity and delivery time. On the other hand, fuzzy TOPSIS ap-
ð29Þ
proach helps to convert DMs’ preference and experience to mean-
ingful results by applying linguistic values to assess each criterion X
n
þ 
and alternative suppliers. Integration with MCGP enables to assign timei X i  d2 þ d2  g 2;min or
order quantities to each supplier by considering the total value i¼1

created from the procurement. According to Chang (2007), MCGP 6 g 2;max ðDelivery time constraintÞ ð30Þ
allows DMs to set multi-choice aspiration levels (MCAL) for each
goal (i.e., one goal mapping multiple aspiration levels) to avoid X
n
þ 
underestimation or overestimation of decision making. The MCAL X i  d4 þ d4 ¼ D ðDemand constraintÞ ð31Þ
on a single attribute can be seen in the following three scenarios i¼1

(Chang, 2008): X i 6 Capi ðSuppliers capacity constraintsÞ ð32Þ

þ 
Xi; di ; di P 0; i ¼ 1; 2; . . . ; n;
(1) As a type of uncertainly/imprecision where a DM considers X 2 F ðF is a feasible set; X is unrestricted in signÞ
only that his/her aspiration level lies in a certain range, in
which the aspiration level can be represented by a range of where CC i is the closeness coefficient of the ith supplier, and pricei is
interval values. the sale price of the ith supplier, and timei is the delivery time level
(2) These aspiration levels indicate different levels of optimism, of the ith supplier. Then D represent the total purchase from X i , and
which can happen when maximizing or minimizing attri- Capi is the capacity of the ith supplier. Other variables are defined as
bute values. For example, a DM who claims ‘‘I expect nothing in MCGP.
less than 95, would be content with 115, and would be very
happy with 130’’. 5. Case study
(3) Conservative policies are usually adopted by the decision
maker to avoid a negative effect. For example, the DM may The company Formosa Watch Co., Ltd. (FWCL) is a large, well-
claim, ‘‘Under company’s resource limitations and the known manufacturer that sells watches in its own chain stores in
incompleteness of available information, I suggest that 120 Asia. For developing new products, its board of directors wishes
million should be our initial aspiration levels for the goal this to select material suppliers to purchase key components in order
year, then, in the future for higher the aspiration levels. It to achieve the competitive advantage in the market. A decision
will be in the long run if we have more available resources’’. committee including three DMs (D1, D2, D1) has been formed to se-
lect a supplier from four qualified suppliers ðS1 ; S2 ; S3 ; S4 Þ. From a
The algorithm of the multi-person multi-criteria decision-marking complete set of criteria, FWCL chooses five supplier selection crite-
with fuzzy TOPSIS and MCGP method for dealing with the supplier ria for the present case:
selection is given as follows:
Step 1. Choose the appropriate linguistic variables for the impor- (1) Relationship closeness (C1).
tance weight of selection criteria and the linguistic ratings (2) Quality of product (C2).
for suppliers. (3) Delivery capabilities (C3).
Step 2. Aggregate the weight w ~ j of criterion C j and pool the DMs’ (4) Warranty level (C4).
ratings to get the aggregated fuzzy rating ~ xij of supplier Si (5) Experience time (C5).
under criterion C j .
Step 3. Construct the fuzzy-decision matrix and normalize the In this study, the hierarchy structure of the decision problem is
matrix. shown in Fig. 5. The integrated fuzzy TOPSIS and MCGP method is
 C.-N. Liao, H.-P. Kao / Expert Systems with Applications 38 (2011) 10803–10811 10809

Selection the Step 4. Table 4 is used to construct a normalized fuzzy-decision

supplier matrix. Using the normalized fuzzy decision matrix in
Table 5, a weighted normalized fuzzy decision matrix is
constructed as shown in Table 6.
Step 5. FPIS and FNIS are determined as follows:

C1 C2 C3 C4 C5
Relationship Quality of Delivery Warranty Experience
S ¼ ½ð0:9; 0:9; 0:9; 0:9; Þ; ð0:9; 0:9; 0:9; 0:9Þ; ð1; 1; 1; 1Þ;
closeness product capabilities level time ð1; 1; 1; 1Þ; ð0:9; 0:9; 0:9; 0:9Þ;

S ¼ ½ð0:35; 0:35; 0:35; 0:35; Þ; ð0:35; 0:35; 0:35; 0:35Þ;

ð0:35; 0:35; 0:35; 0:35Þ; ð0:49; 0:49; 0:49; 0:49Þ;
ð0:35; 0:35; 0:35; 0:35Þ:
MCGP S1 S2 S3 S4
Step 6. Calculate the distance of each supplier from FPIS and FNIS
with respect to each criterion, respectively, as shown in
Fig. 5. Hierarchy structure of decision problem. Tables 7 and 8.
Step 7. Calculate the closeness coefficients of each supplier, as
per Table 9: CC 1 ¼ 0:558; CC 2 ¼ 0:502; CC 3 ¼ 0:516;
applied to solve this problem, and the computational procedure is CC 4 ¼ 0:476.
summarized as follows: Step 8. According to the closeness coefficients (CC i ; i ¼ 1; 2; 3; 4Þ
obtained from Step 7 for each supplier, build the MCGP
Step 1. Three DMs use the linguistic variables shown in Fig. 3 to model to identify the best suppliers and optimum order
assess the importance weight of each supplier criterion; qualities. Similar to Ghodsypour and O’Brien (2001) and
the results of the weights are presented in Table 2. Guneri, Yucel, and Ayyildiz (2009), supplier weights (or
Step 2. Three DMs use the linguistic variables shown in Fig. 4 to priority values) are used as closeness coefficients in an
rate suppliers with respect to each criterion; the results objective function to allocate order quantities among sup-
of the ratings are shown in Table 3. pliers such that the total value of procurement (TVP) is
Step 3. The linguistic evaluations shown in Tables 2 and 3 are con- maximized.
verted into trapezoidal fuzzy numbers to construct a
fuzzy-decision matrix and determine the fuzzy weight of According to the sales record in the last 5 years and the sales
each criterion, as shown in Table 4. forecast by FWCL, the CEO and top managers of FWCL have estab-
lished four goals as follows:
Table 2
Importance weights of criteria for three DMs. (1) The TVP of at least 3500 units from procurement; and the
more the better.
D1 D2 D3
(2) The total cost of procurement of less than 53,200 thousand
C1 H H H
dollars; and the less the better.
C2 H H H
C3 VH VH VH (3) For achieving the procurement levels, the delivery time (per
C4 VH VH H batch) from supplier is set between 4 and 7 days; the less the
C5 H H H better.
(4) For seeking differentiation strategy (i.e., quality leadership),
maintain the current procurement level of less than 5000
units.
Table 3
Ratings of four candidates by DMs according to five criteria.
In addition, the coefficients of variables in model are given by
Criteria Suppliers Decision makers (DMs)
FWCL’s database calculated from the last 5 years record. The unit
D1 D2 D3 material cost for suppliers S1 , S2 , S3 , and S4 are $12, $9, $15, and
C1 S1 G G G $6, respectively, and the capacities of the four candidate suppliers
S2 MG MG G S1 , S2 , S3 , and S4 are 2700, 3500, 2300, and 3100 units, respectively.
S3 VG VG G Furthermore, the delivery time levels of the four candidate suppli-
S4 G G G
ers are 2.5, 4, 6, and 3 days, respectively. The functions and param-
C2 S1 G G VG eters related to FWCL’s supplier selection problem are listed
S2 G G G
S3 VG VG VG
below:
S4 G MG VG
C3 S1 VG VG MG f1 ðXÞ ¼ 0:558x1 þ 0:502x2 þ 0:516x3 þ 0:476x4
S2 G G G  3500 ðTVP goal; the more the betterÞ
S3 MG MG MG
S4 MG MG G
f2 ðXÞ ¼ 12x1 þ 9x2 þ 15x3 þ 6x4  46000 and
C4 S1 VG VG VG
S2 G MG VG 53200 ðprocurement cost goal; the less the betterÞ
S3 VG G G
S4 G G G
f3 ðXÞ ¼ 2:5x1 þ 4x2 þ 6x3 þ 3x4  4 and
C5 S1 VG VG G
S2 G G G 7 ðdelivery time goal; the less the betterÞ
S3 MG MG G
S4 MG MG G
f4 ðXÞ ¼ x1 þ x2 þ x3 þ x4 5000 ðprocurement level goalÞ:
10810 C.-N. Liao, H.-P. Kao / Expert Systems with Applications 38 (2011) 10803–10811

Table 4
Fuzzy decision-matrix and fuzzy weights of four candidates.

C1 C2 C3 C4 C5
S1 (7, 8, 8, 9) (5, 7, 8, 10) (5, 8, 9, 10) (8, 9, 10, 10) (7, 8.7, 9.3, 10)
S2 (5, 6.7, 7.3, 9) (7, 8, 8, 9) (7, 8, 8, 9) (7, 8, 8, 9) (7, 8, 8, 9)
S3 (8, 8.7, 9.3, 10) (8, 9, 10, 10) (5, 6, 7, 8) (7, 8.3, 8.7, 10) (5, 6, 7, 8)
S4 (7, 8, 8, 9) (5, 7.7, 8.3, 10) (5, 6.7, 7.3, 9) (7, 8, 8, 9) (5, 6.7, 7.3, 9)
Weights (0.7, 0.8, 0.8, 0.9) (0.7, 0.8, 0.8, 0.9) (0.7, 0.9, 1.0, 1.0) (0.7, 0.87, 0.93, 1.0) (0.7, 0.8, 0.8, 0.9)

Table 5
Normalized fuzzy decision-matrix.

C1 C2 C3 C4 C5
S1 (0.7, 0.8, 0.8, 0.9) (0.5, 0.7, 0.8, 1) (0.5, 0.8, 0.9, 1) (0.8, 0.9, 1, 1) (0.7, 0.87, 0.93, 1)
S2 (0.5, 0.67, 0.73, 0.9) (0.7, 0.8, 0.8, 0.9) (0.7, 0.8, 0.8, 0.9) (0.7, 0.8, 0.8, 0.9) (0.7, 0.8, 0.8, 0.9)
S3 (0.8, 0.87, 0.93, 1) (0.8, 0.9, 1, 1) (0.5, 0.6, 0.7, 0.8) (0.7, 0.83, 0.87, 1) (0.5, 0.6, 0.7, 0.8)
S4 (0.7, 0.8, 0.8, 0.9) (0.5, 0.77, 0.83, 1) (0.5, 0.67, 0.73, 0.9) (0.7, 0.8, 0.8, 0.9) (0.5, 0.67, 0.73, 0.9)

Table 6
Weighted normalized fuzzy decision-matrix.

C1 C2 C3 C4 C5
S1 (0.49, 0.64, 0.64, 0.81) (0.35, 0.56, 0.64, 0.9) (0.35, 0.72, 0.9, 1) (0.56, 0.78, 0.93, 1) (0.49, 0.69, 0.75, 0.9)
S2 (0.35, 0.53, 0.59, 0.81) (0.49, 0.64, 0.64, 0.81) (0.49, 0.72, 0.8, 0.9) (0.49, 0.69, 0.75, 0.9) (0.49, 0.64, 0.64, 0.81)
S3 (0.56, 0.69, 0.75, 0.9) (0.56, 0.72, 0.8, 0.9) (0.35, 0.54, 0.7, 0.8) (0.49, 0.72, 0.81, 1) (0.35, 0.48, 0.56, 0.72)
S4 (0.49, 0.64, 0.64, 0.81) (0.35, 0.61, 0.67, 0.9) (0.35, 0.6, 0.73, 0.9) (0.49, 0.69, 0.75, 0.9) (0.35, 0.53, 0.59, 0.81)

þ  þ  þ  þ 
Table 7 Min z ¼ d1 þ d1 þ d2 þ d2 þ d3 þ d3 þ d4 þ d4 þ eþ1 þ e1 þ eþ2 þ e2 þ eþ3 þ e3
Distances between FPIS and supplier ratings. þ 
s:t: 0:558x1 þ 0:502x2 þ 0:516x3 þ 0:476x4  d1 þ d1 P 3500
C1 C2 C3 C4 C5 ðTVP goal; the more the betterÞ
þ 
dðS1 ; S Þ 0.28 0.35 0.36 0.25 0.24 12x1 þ 9x2 þ 15x3 þ 6x4  d2 þ d2 ¼ y1
dðS2 ; S Þ 0.37 0.28 0.31 0.33 0.28
ðprocurement cost; the less the betterÞ
dðS3 ; S Þ 0.21 0.20 0.44 0.31 0.40
dðS4 ; S Þ 0.28 0.33 0.41 0.33 0.37 y1  eþ1 þ e1 ¼ 46000 for jy1  g 1;min j
46000 y1 53200 for bound of y1
þ 
2:5x1 þ 4x2 þ 6x3 þ 3x4  d3 þ d3 ¼ y2 for delivery time goal
y2  eþ2 þ e2 ¼ 4 for jy2  g 2;min j
4 6 y2 7 for bound of y2
þ 
Table 8 x1 þ x2 þ x3 þ x4  d4 þ d4 5000 for procurement level
Distances between FNIS and supplier ratings. x1 2700 for the capacity bound of S1
C1 C2 C3 C4 C5 x2 3500 for the capacity bound of S2
x3 2300 for the capacity bound of S3
dðS1 ; S Þ 0.32 0.33 0.46 0.37 0.39
dðS2 ; S Þ 0.27 0.32 0.41 0.26 0.32 x4 3100 for the capacity bound of S4
dðS3 ; S Þ 0.39 0.41 0.30 0.32 0.22 xi P 0; i ¼ 1; 2; 3; 4;
dðS4 ; S Þ 0.32 0.34 0.36 0.26 0.27 þ  þ  þ  þ 
d1 ; d1 ; d2 ; d2 ; d3 ; d3 ; d4 ; d4 ; eþ1 ; e1 ; eþ2 ; e2  0:

where all decision variable coefficients are defined as in FTM model.

These models can be solved using LINGO (Schrage, 2002) to
Table 9 obtain optimal solutions. The best suppliers and their optimum
 
Computation of di , di and CC i . quantities are calculated as follows: S1 ðx1 ¼ 2700Þ; S3 ðx3 ¼ 907Þ;
di

di
 
di þ di

CC i S2 ðx2 ¼ 0Þ; and S4 ðx4 ¼ 0Þ with TVP = 8608.82.
S1 1.86 1.48 3.34 0.558
S2 1.58 1.57 3.14 0.502 6. Conclusion
S3 1.65 1.55 3.20 0.516
S4 1.55 1.71 3.27 0.476
Supplier selection is one of the critical decision-making activi-
ties for firms to obtain competitive advantage. To achieve this goal,
DMs should apply an effective method and select suitable criteria
for supplier selection. Taking both tangible and intangible criteria
Using an integrated fuzzy TOPSIS and MCGP approach, this into account, this paper proposed a novel method, which integrates
problem can be formulated as follows: fuzzy TOPSIS and MCGP, to evaluate suppliers.
 C.-N. Liao, H.-P. Kao / Expert Systems with Applications 38 (2011) 10803–10811
In a decision-making process, the use of linguistic variables in
decision problems is highly beneficial when performance values
cannot be expressed by means of numerical values. In general, sup-
plier evaluation and selection problems are vague and uncertain,
and so fuzzy set theory helps to convert DM preferences and expe-
riences into meaningful results by applying linguistic values to
measure each criterion with respect to every supplier. Employing
MCGP enables us to assign order quantities to each supplier and
thus maximize the total value of procurement. Given that many
multi-choice aspiration levels may exist, a multiple choice method
is most appropriate for this type of decision-making. In addition,
this integrated method allows for the vague aspirations of DMs
to set multiple aspiration levels for supplier selection problems.
The integrated advantage of this method is that it allow for the va-
gue aspirations of DMs to set multiple aspiration levels for supplier
selection problems in which ‘‘the more/higher is better’’ (e.g., ben-
efit criteria) or ‘‘the less/lower is better’’ (e.g., cost criteria).
Furthermore, the proposed method may be useful for various
MCDM problems, such as management problems (e.g., project
management and location selection) and marketing problems
(e.g., new products development and promotion activities) when
available data are inexact, vague, imprecise and uncertain by
nature.
References
Buckley, J. J. (1985). Fuzzy hierarchical analysis. Fuzzy Sets and Systems, 17, 233–247.
Chang, C. T. (2007). Multi-choice goal programming. Omega, 35(4), 389–396.
Chang, C. T. (2008). Revised multi-choice goal programming. Applied Mathematical
Modelling, 32, 2587–2595.
Charnes, A., & Cooper, W. W. (1961). Management model and industrial application of
linear programming (Vol. 1). New York: Wiley.
Chen, C. T. (2000). Extensions of the TOPSIS for group decision-making under fuzzy
environment. Fuzzy Sets and Systems, 114, 1–9.
Chen, C. T., Lin, C. T., & Huang, S. F. (2006). A fuzzy approach for suppler evaluation
and selection in supply chain management. International Journal of Production
Economics, 102, 289–301.
Dickson, G. W. (1966). An analysis of supplier selection system and decision. Journal
of Purchasing, 2(1), 5–17.
Ellram, L. (1990). The supplier selection decision in strategic partnerships. Journal of
Purchasing and Material Management, 26(1), 8–14.
Evans, R. H. (1980). Choice criteria revisited. Journal of Marketing, 44(1), 55–56.
Faez, F., Ghodsypour, S. H., & O’brien, C. (2009). Vendor selection and order
allocation using an integrated fuzzy case-based reasoning and mathematical
programming model. International Journal of Production Economics, 121(2),
395–408.
Ghodsypour, S. H., & O’Brien, C. (1998). A decision support system for supplier
selection using an integrated analytic hierarchy process and linear
programming. International Journal of Production Economics, 56–57, 199–212.
10811
Ghodsypour, S. H., & O’Brien, C. (2001). The total cost of logistics in supplier
selection, under conditions of multiple sourcing, multiple criteria and capacity
constraint. International Journal of Production Economics, 73, 15–27.
Guneri, A. F., Yucel, A., & Ayyildiz, G. (2009). An integrated fuzzy-lp approach for a
supplier selection problem in supply chain management. Expert Systems with
Applications, 36(5), 9223–9228.
Ha, S. H., & Krishnan, R. (2008). A hybrid approach to supplier selection problem for
the maintenance of a competitive supply chain. Expert Systems with Applications,
34(2), 1303–1311.
Ignizio, J. P. (1976). Goal programming and extensions. Lexington, MA: Lexington
Books.
Kahraman, C., Cevik, S., Ates, N. Y., & Güfer, M. (2007). Fuzzy multi-criteria
evaluation robotic systems. Computers & Industrial Engineering, 52, 414–433.
Kokangul, A., & Susuz, Z. (2009). Integrated analytical hierarch process and
mathematical programming to supplier selection problem with quality
discount. Applied Mathematical Modelling, 33, 1417–1429.
Krajewsld, L. J., & Ritzman, L. P. (1996). Operations management strategy and analysis.
London: Addison-Wesley Publishing Co..
Kumar, M., Vrat, P., & Shankar, R. (2006). A fuzzy programming approach for vendor
selection problem in a supply chain. International Journal of Production
Economics, 101, 273–285.
Liao, C. N. (2009). Formulating the multi-segment goal programming. Computers &
Industrial Engineering, 56, 138–141.
Lin, H. T., & Chang, W. L. (2008). Order selection and pricing methods using flexible
quantity and fuzzy approach for buyer evaluation. European Journal of
operational Research, 187(2), 415–428.
Önüt, S., Kara, S. S., & Isik, E. (2009). Long term supplier selection using a combined
fuzzy MCDM approach: A case study for a telecommunication company. Expert
Systems with Applications, 36(2), 3887–3895.
Pi, W. N., & Low, C. (2005). Supplier evaluation and selection using Taguchi loss
functions. International Journal of Advanced Manufacturing Technology, 26,
155–160.
Romero, C. (2001). Extended lexicographic goal programming: A unifying approach.
Omega, 29(1), 63–71.
Sarami, M., Mousavi, S. F., & Sanayei, A. (2009). TQM consultant selection in SMEs
with TOPSIS under fuzzy environment. Expert Systems with Applications, 36(2),
2742–2749.
Schrage, L. (2002). LINGO Release 8.0. LINDO System, Inc..
Shin, H., Collier, D. A., & Wilson, D. D. (2000). Supplier management orientations and
supplier/buyer performance. Journal of Operations Management, 18, 317–333.
Shipley, D. D. (1985). Reseller’s supplier selection criteria for different consumer
products. European Journal of Marketing, 19(7), 26–36.
Tam, M. C. Y., & Tummala, V. M. R. (2001). An application of the AHP in vendor
selection of a telecommunications system. Omega, 29, 171–182.
Tamiz, M., Jones, D., & Romero, C. (1998). Goal programming for decision marking:
An overview of the current state-of-the-art. European Journal of Operational
Research, 111(3), 567–581.
Wang, J. W., Cheng, C. H., & Cheng, H. K. (2009). Fuzzy hierarchical TOPSIS for
supplier selection. Applied Soft Computing, 9, 377–386.
Weber, C. L., Current, J. R., & Benton, W. C. (1991). Vendor selection criteria and
methods. European Journal of Operational Research, 50(1), 2–18.
Zadeh, L. (1965). Fuzzy sets. Information Control, 8, 338–353.
Zimmermann, H. J. (1978). Fuzzy programming and linear programming with
several objective functions. Fuzzy Sets and Systems, 1, 45–55.