Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

Expert Systems with Applications 38 (2011) 9773–9779

Contents lists available at ScienceDirect

Expert Systems with Applications


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

Comparative analysis of multi-criteria decision making methodologies


and implementation of a warehouse location selection problem
Tuncay Özcan ⇑, Numan Çelebi, Sß akir Esnaf
Istanbul University, Department of Industrial Engineering, 34320 Avcilar, Istanbul, Turkey

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

Keywords: For the solution of decision making problems with multi criteria, the literature presents many method-
Multi criteria decision making ologies under the title of decision theory. In this context, AHP, TOPSIS, ELECTRE and Grey Theory are well-
Warehouse location selection known and the most acceptable methodologies. Firstly, in this study; these methodologies are compared
TOPSIS in terms of main characteristic of decision theory and thus advantages and disadvantages of these meth-
ELECTRE
odologies are offered. Later, the application of these methodologies on the warehouse selection problem,
Grey Theory
which is one of the main topics of logistics management that has a wide range of applications with multi-
criteria decision making methodologies, is presented as a case study which is characterized in retail sec-
tor, that maintains high uncertainity and product variety and then how to choose the best warehouse
location among many alternatives has been shown.
Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction Ang, 1999) and water resources planning (Anand, 1995) that have
increasing importance in recent years.
Multi-criteria decision making can be defined as the evaluation Unlike these methods, Grey Theory, extremely high mathemat-
of the alternatives for the purpose of selection or ranking, using a ical analysis of the systems that are partly known and partly
number of qualitative and/or quantitative criteria that have differ- unknown and defined as ‘‘weak knowledge’’ and ‘‘insufficient
ent measurement units. To this end, in the context of decision the- data’’, was first introduced by Deng (1982). Grey Theory examines
ory; ELECTRE, AHP and TOPSIS can be sorted as primary methods. the interactional analysis when the decision-making process is not
Depending on the structure of decision problems, the applica- clear, there are a great number of input data and it is discrete and
tion of this method is found in a wide area. For example, TOPSIS insufficient (Wen, 2004). In recent years, Grey Theory has been
method, that has chosen the best alternative based on a maximiza- successfully applied in many decision making problems, such as
tion of the distance from the negative ideal point and minimization supplier selection (Li, Yamaguchi, & Nagai, 2006), facility layout
of the distance from the positive ideal point, was not only applied selection (Kuo, Yang, & Huang, 2008), financial performance evalu-
to areas such as performance evaluation with the use of financial ation (Kung & Wen, 2007), demand forecasting (Wang, 2004) and
investment decisions (Kim, Park, & Yoon, 1997) and financial ratios material selection (Chan & Tong, 2007).
(Deng, Yeh, & Willis, 2000; Feng & Wang, 2001), but also applied to In this study, it is intended to analyze both conceptual-oriented
problems such as flexible manufacturing systems (Agrawal, Kohli, and application-oriented structure of these methodologies that are
& Gupta, 1991) and selection of production processes (Chau & used in the solution of decision-making problems with multi-
Parkan, 1995; Parkan & Wu, 1995) within the scope of operation criteria. Therefore, the next sections of this study will be in the fol-
management. Similarly, ELECTRE methods (Electre I, IS, II, III, IV, lowing way.
A) that have chosen the best alternative by means of pairwise In the second section, multi-criteria decision making methodol-
comparison of all alternatives; within the decision problems such ogies with a comparative analysis, that is used to provide guidance
as factory location and production center selection (Lopez & to the decision maker depending on the structure of decision prob-
Gonzalez, 2003), especially has been applied to solve the issue of lems, will be presented. In the third section, TOPSIS and ELECTRE
environmental management, environmental valuation (Rogers & method and Grey Theory will be used for the solution in retail
Bruen, 1998), solid waste management system selection (Poh & sector about a case study of the problem of warehouse location
selection which is one of the strategic decision making processes
of logistics management in which decision making methodologies
finds a broad practice area. The findings acquired from the analysis
⇑ Corresponding author. Tel.: +90 212 4737070; fax: +90 212 4737180.
of conceptual and practical structures of multi-criteria decision
E-mail addresses: tuncay.ozcan@gmail.com (T. Özcan), ncelebi@istanbul.edu.tr
(N. Çelebi), sesnaf@istanbul.edu.tr (S
ß . Esnaf).
making methodologies will be discussed in the last section.

0957-4174/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.eswa.2011.02.022
9774 T. Özcan et al. / Expert Systems with Applications 38 (2011) 9773–9779

2. The comparative analysis of decision making methodologies more important than C criterion, the condition that A criterion is
25 times more important than C criterion cannot be explained with
The structure of the methods of AHP, TOPSIS, ELECTRE I, ELEC- AHP. TOPSIS and ELECTRE methodologies do not put forward a spe-
TRE II and ELECTRE III used in multi-criteria decision making can cific method about weight determination. In some decision making
be analysed as follows on the basis of the study of Karagiannidis problems in which these methodologies are used, criterion weights
and Moussiopoulos (1997) in which the comparative analysis of are given without using any calculation. On the other hand, while
the structure of ELECTRE is presented, and the characteristics used determining the weight in these methodologies, it is generally
in the study of Shih, Shyur, and Lee (2007) which includes the com- common to use the linear and vector normalizations (Shih et al.,
parison of AHP and TOPSIS. 2007).
Thecore (main) process: Decision making methodologies are sep- The type and number of outranking relationship: The type and
arated from each other through different calculation methods. The number of the ranking relationships among the criteria used in
steps that are separative from other decisions making methods and the solution of decision making problems differ according to meth-
important in the solution algorithm are named as the core process. odologies. In AHP, all the factors in the all the levels of hierarchy
The core process in AHP is to create hierarchy and pairwise com- are pairly compared to each other. Thus, the pairwise comparison
parison matrices in all the levels of hierarchy (Saaty, 1980). In TOP- number determining the outranking relations in an N-dimensional
SIS method, the calculation of each alternative distance from the comparison matrix is N (N  1)/2. That the number of pairwise
positive ideal and the negative ideal solutions draws attention. comparison matrix can be too many is an important disadvantage
While ELECTRE I and ELECTRE II methods are separated from the of AHP and in the situations like when the number of alternatives
other methods through the determination of concordance and dis- and criteria are a lot, the opportunity of carrying out the method-
cordance matrices for each criterion and alternative pair, ELECTRE ology is substantially prevented. TOPSIS and ELECTRE I methodol-
III method, different from the other methods, is based on the prin- ogies need less input compared to AHP and eliminate the necessity
ciple of fuzzy logic and uses the preference and indifference of comparisons of pairs. While only one outranking relation is put
thresholds while determining the concordance and discordance in- forward while evaluating the criteria and alternatives in TOPSIS
dexes (Roy, 1991). and ELECTRE I methods, two way outranking relations among
The methodology of determining the weight of criteria: In multi- alternatives is put forward in ELECTRE II. In ELECTRE III, there is
criteria decision making methodologies, for the rank among the a outranking relation based on one way fuzzy logic (Karagiannidis
alternatives and the determination of their preference, the neces- & Moussiopoulos, 1997).
sity to determine the relative importance of criteria reveals. Also, The control of consistency: It is an important necessity to mea-
AHP, TOPSIS, ELECTRE I, ELECTRE II, ELECTRE III methods are in sure the consistency of judgements of decision maker in the prob-
need of measuring the relative importance of criteria in the solu- lem of decision making. The consistency of judgements the
tion of the decision problem. In the determination of the weight decision maker uses in AHP while creating pairwise comparison
of criteria, while some parts of methodologies put forward a spe- matrices is measured through calculating the consistency index
cific method, the others suggest a solution which determines the (Saaty, 1980). The limitation of consistency is one of the most
importance of the decision maker based on the number of criteria. important advantages of AHP. The consistency is not controlled
Being separated from the other methodologies through using pair- in TOPSIS, ELECTRE I and ELECTRE II methods. For example, if A
wise comparison matrix in order to determine the importance of alternative is preferred to B alternative and B alternative to C, as
criteria, AHP presents a specific method (Saaty, 1980). However, a consistency principle in AHP, A alternative is preferred to C alter-
while using pairwise comparison matrices, the based on the 1–9 native. On the other hand, there may not be such a preference rela-
scale and the limitation cause by this principle are one of the tionship TOPSIS, ELECTRE I and ELECTRE II. In ELECTRE III method
important disadvantages of AHP. For example, if A criterion is 5 the consistency is provided through indifference and preference
times more important than B criterion and B criterion is 5 times thresholds determined for criteria.

Table 1
Comparative analysis of multi-criteria decision making methodologies.

Characteristics AHP TOPSIS ELECTRE I ELECTRE II ELECTRE III


1. Core process Creating hierarchical Calculating distance to Determining Determining Determining concordance and
structure and pairwise positive and negative concordance and concordance and discordance indexes with
comparison matrices ideal point discordance indexes discordance indexes indifference and preference
thresholds
2. Necessity to Yes Yes Yes Yes Yes
quantify the
relative
importance of
criteria
3. Determining of Pairwise comparison No specific method. No specific method. No specific method. No specific method. Based on
weights matrices. 1–9 scale Linear or vector Based on decision maker Based on decision maker decision maker
normalization
4. Number and type N (N  1)/2 1 1 2 1 Fuzzy
of outranking
relations
5. Consistency Provided None None None Provided
Check
6. Problem Little number of Large number of Large number of Large number of Objective and quantitative data,
Structure alternative and criteria, alternative and criteria, alternative and criteria, alternative and criteria, usage of fuzzy logic
quantitative or objective and objective and objective and
qualitative data quantitative data quantitative data quantitative data
7. Final Results Global, net ordering Global, net ordering A kernel A partial pre-order A partial pre-order

Source: Saaty (1980), Hwang and Yoon (1981), Roy (1991), Karagiannidis and Moussiopoulos (1997), Shih et al. (2007).
T. Özcan et al. / Expert Systems with Applications 38 (2011) 9773–9779 9775

Problem structure: The selection of the correct methodology re- warehouse alternatives, many studies concerning the site selection
lated to the structure decision making problems is really important decision in fuzzy environment also exist in literature (Kuo, Chi, &
for the decision maker. AHP is preferred in the situations when the Kao, 2002; Tabari, Kaboli, Aryanezhad, Shahanaghi, & Siadat,
decision making problem can be dissociated as criteria, subcriteria 2008; Chen, 2001; Kahraman, Ruan, & Doğan, 2003).
and alternatives, and the effect of each subject is demanded for The decision making problem forming the application part of
measuring. Furthermore, since it is necessary to make pairwise the study can be expressed in more specific way. The business
comparisons in all the levels of hierarchy, it gets harder to perform operating in retail sector with demand uncertainty and high prod-
AHP as the number of alternatives and criteria increase. On the uct variety has detected that it will be inadequate in performing
other hand, AHP can be performed easily disregarding the data the current warehouse parallel to the growing rate beginning from
applying evaluation of alternatives based on criteria is quantitative the end of 2008. Thus, in the manner of covering the coming five
or qualitative. TOPSIS method draws attention as its simplicity in year growing target, the business plans to make a decision about
perception and use. TOPSIS and ELECTRE methods can be per- the new storage area which will provide the logistic operations
formed easily when the number of alternatives and criteria is too completely and thoroughly. As a result of the research, four differ-
many. In addition, these methods are more suitable in cases when ent alternative warehouses has been specified. In the decision for
the data is provided as objective and quantitative. warehouse location selection, the evaluation criteria for criteria
Final results: In AHP and TOPSIS methodologies, after complet- are explained as follows in a way that covers the criteria based
ing the method steps, global and net ranking is attained among on capacity, cost and customer related to the structure of sector
alternatives. In ELECTRE I method, the result input acquired is in and business.
the shape of kernel. The preference and ranking relationships Unit price: One of the fundamental determiner of warehouse
among alternatives are determined with this kernel. In ELECTRE location decision is the unit price per m2. Decreasing of unit price
II method, thus, a two way ranking is acquired and the outranking among different warehouse location alternatives causes choice
relations among alternatives are determined through associating priority.
the two way ranking. In ELECTRE III method, as a result of the com- Stock holding capacity: In the business structure with a high
parison of concordance and discordance indexes with concordance product variety, a stock holding capacity which will accord with
and discordance thresholds partial ranking is acquired at the paral- the grow rate is really important. While a warehouse location with
lel of reliability matrix formed among alternatives (Karagiannidis & a low shelf capacity causes a problem for the operational flow to
Moussiopoulos, 1997). operate completely, a high stock holding capacity causes the for-
Basic results attained from the comparative analysis of decision mation of inactive area and the increase in cost of warehouse area.
making methodologies are given in Table 1. On the other hand, Therefore, the warehouse alternatives that are out of top and bot-
Grey Theory is used in the solution of decision making problems tom borders of identified stock holding capacity are eliminated in
related to grey systems defined as a little information and poor the process of decision making.
data known on specific scale and unknown on specific scale (Yuan, Average distance to shops: Decrease in the presentation period of
2007). products to client and supply chain cycle time provide an impor-
tant competition advantage for retail sector businesses. This situa-
3. The case study tion has a vital importance especially as to get ahead of rival
businesses for the products that will present to client for the first
In order to analyse the practice oriented structure along with time, to run out of products when there is stock out of products
conceptual analysis of decision making methodologies, a case with high sale volume and to prevent the cost of sales and prestige
study including the practice of these methodologies will be pre- loss. Minimization of average distance of warehouse location alter-
sented in this section for the problem of warehouse location selec- natives to main shops forming the important part of the sales vol-
tion. Defining the decision making problem, putting forward the ume is one of the basic purpose of the decision maker.
weighting methodology of evaluation criteria and solving the prob- Average distance to main suppliers: Another factor that will re-
lem with TOPSIS and ELECTRE methods, practicing the Grey Theory duce supply chain cycle time is the distance of the warehouse loca-
carried out to eliminate the drawbacks of results acquired with tion to main supplier centers. Minimization of this criterion
these methods compose the main steps of the case study. provides to reduce product presentation time and logistics trans-
portation costs.
3.1. The definition of problem Movement flexibility: The movement flexibility resulting from
the evaluation of architectural and layout factors of warehouse
The problem of warehouse location selection contains an effi- location equals to the indicators like the conformity of warehouse
cient strategic investment decision on long term and business prof- location to different layout arrangement and the useful area rate in
itability. At this point, one of the most important decision making total storage area. Thus alternative warehouse locations are evalu-
process of logistic administrators is the location decision of the dis- ated based on 0–4 scale (really bad, bad, average, good, really
tribution center (Chen, 2001). In the studies dealing with this deci- good).
sion making process, while firstly, a solution based on the cost For the warehouse location selection, four different warehouse
minimization traditionally is recommended (Lee, 1993), after then location alternatives having the acceptable performance in the ba-
there are also studies concerning profit maximization (Hakimi & sis of the criteria defined above were introduced. The hierarchical
Kuo, 1991). In addition to the cost and profitability indicators, un- structure of decision making problem is shown in Fig. 1.
der increasing competition conditions of todays, the indicators The performance values of these four warehouse locations
revealing customer service level and customer satisfaction are also according to defined criteria are given in Table 2.
demanded. In this sense, Korpela and Lehmusvaara (1999) devel-
oped a customer oriented approach to evaluate alternative stocks 3.2. Weighting procedure of criteria
and logistic services based on AHP and mixed integer program-
ming model. For the warehouse location decision, Korpela, Whereas some of multiple criteria decision making methodolo-
Lehmusvaara, and Nisonen (2007) used AHP and DEA methods to- gies introduce a specific way for weighting of criteria, others pro-
gether. While Ho and Emrouznejad (2009) developed an applica- pose a solution that decision makers determine the weightings
tion based on AHP and goal programming in order to evaluate according to the number of criteria. In this study, the Simos
9776 T. Özcan et al. / Expert Systems with Applications 38 (2011) 9773–9779

between two consecutive criteria this problem is solved. The


bigger difference between criteria weightings means more
amount white cards.

By using Simos procedure for warehouse location selection


problem the weighting of unit price (UP), stock holding capacity
(SHC), average distance to shops (DS), average distance to main
suppliers (DMS) and movement flexibility (MF) criteria is calcu-
lated as follows:

 Criteria set F = {UP, SHC, DS, DMS, MF} is formed.


 Decision makers determine the order of the importance among
criteria as MF, DMS, DS, UP and SHC.
 Here, decision makers valuate the average distance to main sup-
plier (DMS) and average distance to shops (DS) as equally
important and the difference of importance among unit price
(UP), DS and DMS is determined to be in the very high level
according to the difference between other two consecutive
criteria.

The decision maker choices which are mentioned above and


calculation steps of weighting criteria in the warehouse location
problem using Simos procedure are shown in Table 3.
In the first step, criteria subsets are formed with same impor-
tant criteria and in the parallel of this relation order and counting
Fig. 1. The decision hierarchy of the warehouse location selection problem.
number for each criterion are appointed. In the second step, aver-
age weighting value for every criteria subset is calculated, finally
the classification value for criteria is proportioned to the sum of
Table 2 consecutive numbers and weighting values are calculated. The
Performance values of warehouse location alternatives. importance weights of criteria in the decision problem are;
Alternative Unit Stock Average Average Movement
{WUP, WSHC, WDS, WDMS, WMF} = {0.29, 0.35, 0.15, 0.15, 0.06}.
price holding distance distance flexibility
($/m2) capacity to shops to main 3.3. Application of TOPSIS method to warehouse selection problem
(unit) (kilometer) suppliers
(kilometer)
TOPSIS is a multiple criteria decision making methodology
Warehouse A 7 100.000 20 14 3 which determines solution alternatives from a finite set in the ba-
Warehouse B 10 120.000 8 10 1
sis of maximizing the distance from the negative ideal point and
Warehouse C 8 150.000 12 12 2
Warehouse D 6 180.000 16 13 4 minimizing the distance from the positive ideal point. (Olson,
2004). TOPSIS is interesting with its need for decision maker’s lim-
ited number of subjective input. Only subjective input is in the cri-
teria weighting phase. The model algorithms steps of TOPSIS
procedure was used for weighting criteria. In this method every (Olson, 2004) and its practice in the case study is as follows:
criteria is matched with a gaming card. For a decision making prob-
lem with n criteria, n card is picked. The process of this method is Step 1. For n criteria performance data of m alternatives are col-
as follows (Figueira & Roy, 2002): lected. Raw values (xij) are transformed to normalized val-
ues (rij).
 Criteria are arranged from the least to the most important. Thus, xij
decision maker classifies criteria according to their importance r ij ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pm 2 i ¼ 1; 2; . . . ; m; j ¼ 1; 2; . . . ; n: ð1Þ
i¼1 xij
from the least to the most important. If some criteria have the
same importance according to the decision maker’s view, a Step 2. For each criterion the importance weights wjare deter-
sub-set of these criteria is established. mined. To determine the weights any technique can be
 In classification, the thought about two consecutive criteria’s used. By the help of weights of criteria, weighted normal-
importance may be closer or less close. This situation means ized values are calculated.
in determining weightings, the difference between consecutive
criteria is very small or big. Therefore, by putting white cards v ij ¼ wj xij i ¼ 1; 2; . . . ; m; j ¼ 1; 2; . . . ; n: ð2Þ

Table 3
Weighting of criteria by using Simos procedure.

Subsets that has equal weight Number of Criteria Positions Non-normalized weights Normalized weights Total

{MF} 1 1 1 (1/17) 100 = 5.88  6 6
{DS, DMS} 2 2, 3 (2 + 3)/2 = 2.5 (2.5/17)⁄100 = 14.70  15 30
White Card 1 (4) ...
{UP} 1 5 5 (5/17)⁄100 = 29.41  29 29
{CSC} 1 6 6 (6/17)⁄100 = 35.29  35 35
Total 6 17 100
T. Özcan et al. / Expert Systems with Applications 38 (2011) 9773–9779 9777

The weighted normalized values that are calculated by 3.4. Application of ELECTRE method to warehouse selection problem
using formula (1) and (2) and criteria weights which are
determined by Simos method are shown in Table 4. The ELECTRE method, having two basic phases as constructing
Step 3. For every criterion ideal alternative with best performance preference relations and the by usage of these relations to achieve
(s+) and worst performance (s) is determined. the final ranking of alternatives, have six different methods as ELEC-
If j is the benefit criteria; TRE I, II, IS, III, IV and A (Karagiannidis & Moussiopoulos, 1997). Dif-
ferent ELECTRE methods show difference at how preference
sþ ¼ fv 1j ; v 2j ; . . . ; v mj g ¼ fmax v 1j for 8 j 2 ng: ð3Þ relations between alternatives obtained and how to achieve rank-
ing of solution by using these relations (Wang & Triantaphyllou,
If j is the cost criteria;
2008).
sþ ¼ fv 1j ; v 2j ; . . . ; v mj g ¼ fmin v 1j for 8 j 2 ng; ð4Þ If ELECTRE I method applied to warehouse location selection
problem; like TOPSIS method’s first two steps, weighted normal-

s values are determined as a benefit criteria by selecting ized matrix is obtained by using criteria weightings obtained from
the minimum value and as a cost criteria by selecting the the exercise of Simos procedure. ELECTRE methods are based on
maximum value. evaluation of concordance and discordance indexes. In the alterna-
In this problem the values that shows best performed ideal tive pair of a and b, the concordance index of a measures the
alternative and worst performed values for every criteria strength of ‘‘alternative a is at least as good as alternative b’’
are; hypothesis. For every alternative pair (a, b) concordance index C
(a, b) is calculated as follows;
sþ ¼ ð0:11; 0:22; 0:04; 0:06; 0:04Þ; P
s ¼ ð0:18; 0:12; 0:10; 0:09; 0:01Þ: i2Q ða;bÞ wi
Cða; bÞ ¼ Pm : ð8Þ
  i1 wi
Step 4. Every alternative’s distance to the best alternative Dþ
  i
and worst alternative Di for all criteria is calculated. Q (a, b) expresses the criteria set where a is equal to or chosen in-
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi stead of b. In the next step, for every (a, b) alternative pair, discor-
uX
u n  2 dance index D (a, b) is calculated as follows;
Di ¼ t
þ
v ij  sþj for i ¼ 1; 2; . . . ; m; ð5Þ
j¼1 maxðg i ðbÞ  g i ðaÞÞ
i
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Dða; bÞ ¼ : ð9Þ
uX max jg i ðbÞ  g i ðaÞj
u n  2
Di ¼ t

v ij  sj for i ¼ 1; 2; . . . ; m: ð6Þ
j¼1
If formulas (8) and (9) are used for warehouse location problem,
concordance and discordance indexes are calculated and obtained
values are given in Table 5.
Using (5) and (6) formulas for the warehouse location alter- For every alternative pair (a, b) concordance threshold (Cavr) and
natives, the distance values to the positive and negative discordance threshold (Davr) is calculated. These values for ware-
solutions are respectively; house location selection problem are calculated as; Cavr = 0.50
and Davr = 0.67.
Dþ ¼ ð0:14; 0:08; 0:06; 0:05Þ and In the last step, according to the condition if C (a, b) P Cavr and D
D ¼ ð0:02; 0:09; 0:09; 0:13Þ: (a, b) 6 Davr, alternative A is chosen instead of alternative B, outran-
king relations between alternatives are determined. The outran-
Step 5. For every alternative by dividing distance to the negative
king relations matrix for warehouse location selection problem is
solution by the sum of distance to the positive and nega-
shown in Table 6
tive solution Ci is determined. Ci exhibits the similarity to
According to this matrix; final result would be Warehouse
the positive ideal solution. According to the magnitude
D > Warehouse B = Warehouse C > Warehouse A. Warehouse D is
of Ci alternatives are arranged. The biggest Ci value is
the best alternative in the ELECTRE method like as the TOPSIS
selected.
method.
Di
Ci ¼ i ¼ 1; 2; . . . ; m and 0 6 C i 6 1: ð7Þ
Di þ Dþi 3.5. Application of Grey Theory to warehouse selection problem

The Ci values that exhibit the similarity to positive ideal solution for In several decision problems, performance indicator of the crite-
warehouse location alternatives are; C i ¼ ð0:15; 0:53; 0:60; 0:74Þ.In ria used for evaluation of alternatives has the purpose of approxi-
the parallel of these values Warehouse D is the best for warehouse mation to a certain reference value other than maximization or
location according to the TOPSIS method for the company. A rank- minimization. For example, in the warehouse location problem,
ing is obtained between warehouses as Warehouse D, Warehouse as regards stock holding capacity criteria, the warehouse alterna-
C, Warehouse B and Warehouse A. tive with low stock loading capacity creates problems in healthy
running of operational flow; high loading capacity, on the other
hand, causes idle area and higher warehouse area costs. Optimum
Table 4 value determined for this criterion is a warehouse with 140.000
Determining weighted normalized values of warehouse location alternatives.

Unit Stock Average Average Movement Table 5


price holding distance distance flexibility The concordance and discordance indexes of warehouse alternatives.
capacity to shops to main
suppliers C (a, b) A B C D D (a,b) A B C D
Warehouse A 0.18 0.12 0.10 0.09 0.03 A 1 0.06 0.06 0 A 0 1 1 1
Warehouse B 0.13 0.15 0.04 0.06 0.01 B 0.94 1 0.59 0.3 B 0.33 0 1 1
Warehouse C 0.15 0.19 0.06 0.07 0.02 C 0.94 0.41 1 0.3 C 0.14 0.5 0 1
Warehouse D 0.11 0.22 0.08 0.08 0.04 D 1 0.7 0.7 1 D 0 0.57 0.5 0
9778 T. Özcan et al. / Expert Systems with Applications 38 (2011) 9773–9779

Table 6 Table 8
Outranking relations matrix for warehouse alternatives. Calculation of difference values for warehouse alternatives.

A B C D Unit Stock Average Average distance Movement


price holding distance to to main flexibility
A –
capacity shops suppliers
B ⁄ –
C ⁄ – DA 0.25 1 1 1 0.34
D ⁄ ⁄ ⁄ – DB 1 0.5 0 0 1
DC 0.6 0.25 0.25 0.5 0.67
DD 0 1 0.67 0.75 0
DMIN 0 0.25 0 0 0
pieces of stock capacity; whereas lower and upper limits are DMAX 1 1 1 1 1
120.000 and 160.000 pieces.
An alternative solution method is being offered with Grey The-
ory in order to correct the deficiencies caused by evaluating perfor- Table 9
mance indicators differently when applying TOPSIS and ELECTRE Calculation of grey relational coefficient and grey relational grade for warehouse
methods. At this point, Grey Theory is an effective solution meth- alternatives.
odology for analysis of mathematical systems characterized by Unit Stock Average Average Movement Grey
uncertain information and insufficient data (Li et al.,, 2006). The price holding distance distance to flexibility relational
steps of this solution method which uses different normalization capacity to shops main grade (ri)
formulas and the grey theory procedure depending on the direc- suppliers

tion of purpose function of performance indicators is given below c(A) 0.667 0.5 0.333 0.333 0.595 0.504
(Yuan, 2007): c(B) 0.333 0.75 1 1 0.333 0.679
c(C) 0.455 1 0.667 0.5 0.427 0.682
c(D) 1 0.5 0.427 0.4 1 0.649
Step 1. Reference data set X0 = (d01, d02, . . . , d0m) is created. X0 ref-
erence data consists of the m value which displays the
most wanted values for criteria. Reference data set deter-
mined for the problem selecting store location is Step 3. The difference between data sets is calculated with
X0 = {XUP, XSHC, XDS, XDMS, XMF} = {5, 140000, 0, 0, 4}. Di = (jd01  di1j, jd02  di2j, . . . , jd0m  dimj) as well as the
Step 2. Xi = (di1, di2, . . . , dim) comparison data is determined which global maximum (Dmax) and global minimum (Dmin)
shows the performance values of each alternative against values between data set difference values for each crite-
the criteria, where i = 1, 2, . . . , k defines the alternative rion. The values for the problem of warehouse location
number; each comparison data set includes an m value. selection are given in Table 8.
Depending on the structure of criteria, performance of Step 4. Each data point in data difference sets is transformed into
the alternatives is standardized using the following for- grey relational coefficient. Grey relational coefficient of the
mula: data point ‘‘j’’ in difference set ‘‘i’’ is calculated with the
In performance indicators of the criteria which are aimed following formula:
for maximization; i.e. income, benefit etc. D min þnD max
ci ðjÞ ¼ : ð13Þ
X i ðkÞ  min X i ðkÞ Di ðjÞ þ nD max
X i ðkÞ ¼ : ð10Þ
max X i ðkÞ  min X i ðkÞ
In performance indicators of the criteria which are aimed Di(j), Di is the j. value in the difference set. Coefficient n is a
for minimization; i.e. cost, distance etc. value between 0 and 1, and is used to decrease the effect of
Dmax, which is the extreme value in the data set. This coef-
max X i ðkÞ  X i ðkÞ ficient is taken as 0.5 in most problems.
X i ðkÞ ¼ : ð11Þ
max X i ðkÞ  min X i ðkÞ Step 5. Grey relational grade of each data set, therefore, each
alternative is calculated. ri, which is the grey relational
In performance indicators which are aimed to be close to a
grade of alternative i, is calculated as follows:
certain optimum value;
Where Ui, is the optimum performance value for criteria i; X
m
ri ¼ ðci ðnÞ wðnÞÞ: ð14Þ
jX i ðkÞ  U i j n¼1
X i ðkÞ ¼ 1  : ð12Þ
max jX i ðkÞ  U i j
For warehouse location alternatives, grey relational coefficient and
In Grey Theory application, criteria weights determined by grey relational grade can be calculated by using formulas (13)–(14),
Simos method shall be used. Normalized data created by as shown in Table 9.
using formulas (10)–(12) are presented in Table 7. When the criteria are ranked according to their grey relational
grade, the priority ranking is obtained as Warehouse C > Warehouse
Table 7 B > Warehouse D > Warehouse A. According to the Grey Theory, the
Creation of normalized data for warehouse alternatives. best alternative for warehouse location is Warehouse C.
Unit Stock Average Average distance Movement
price holding distance to to main suppliers flexibility 4. Conclusions
capacity shops
X0 1 1 1 1 1 Multi criteria decision-making methodologies find application
A 0.75 0 0 0 0.66 in a wide area. Several methodologies developed for this purpose
B 0 0.5 1 1 0 are employed according to the structure of decision problem and
C 0.4 0.75 0.75 0.5 0.33 the preferences of decision maker. Firstly, in this study; compara-
D 1 0 0.33 0.25 1
MIN OPT MIN MIN MAX
tive analysis of the three most widely used basic methodologies,
namely AHP, TOPSIS and ELECTRE has been conducted and basic
T. Özcan et al. / Expert Systems with Applications 38 (2011) 9773–9779 9779

Table 10 Hakimi, S. L., & Kuo, C. C. (1991). On a general network location allocation problem.
Preference ranking of warehouse alternatives depending on decision methodologies. European Journal of Operational Research, 108, 135–142.
Ho, W., & Emrouznejad, A. (2009). Multi-criteria logistics distribution network
TOPSIS ELECTRE Grey Theory design using SAS/OR. Expert Systems with Applications, 36, 7288–7298.
Hwang, C. L., & Yoon, K. (1981). Multiple Attribute Decision Making: Methods and
Warehouse A 4 3 4
Applications. New-York: Springer-Verlag.
Warehouse B 3 2 2
Kahraman, C., Ruan, D., & Doğan, I. (2003). Fuzzy group decision-making for facility
Warehouse C 2 2 1 location selection. Information Sciences, 157, 135–153.
Warehouse D 1 1 3 Karagiannidis, A., & Moussiopoulos, N. (1997). Application of ELECTRE III for the
integrated management of municipal solid wastes in the Greater Athens Area.
European Journal of Operational Research, 97, 439–449.
Kim, G., Park, C., & Yoon, K. P. (1997). Identifying investment opportunities for
characteristics of these methods have been displayed. Then, the advanced manufacturing systems with comparative-integrated performance
problem of selecting warehouse location, is one of the basic issues measurement. International Journal of Production Economics, 50, 23–33.
Korpela, J., & Lehmusvaara, A. (1999). A customer oriented approach to warehouse
of logistics management, which is one area where multi criteria network evaluation and design. International Journal of Production Economics, 59,
decision making is most widely used has been elaborated. Ware- 135–146.
house location problem is examined with the case study in retail Korpela, J., Lehmusvaara, A., & Nisonen, J. (2007). Warehouse operator selection by
combining AHP and DEA methodologies. International Journal of Production
sector. TOPSIS and ELECTRE methodologies are applied for solution Economics, 108, 135–142.
of this problem. However, when using these methods, criteria have Kung, C., & Wen, K. (2007). Applying grey relational analysis and grey decision-
been evaluated parallel to two basic purposes of maximization and making to evaluate the relationship between company attributes and its
financial performance – A case study of venture capital enterprises in Taiwan.
minimization; the fact has been ignored that the criteria like stock
Decision Support Systems, 43, 842–852.
holding capacity had to define a certain lower limit, an optimum Kuo, R. J., Chi, S. C., & Kao, S. S. (2002). A decision support system for selecting
value and an upper limit. At this point, Grey Theory was applied convenience store location through integration of fuzzy AHP and artificial
neural network. Computers in Industry, 47, 199–214.
to the problem of selecting warehouse location for the purpose
Kuo, Y., Yang, T., & Huang, G. (2008). The use of grey relational analysis in solving
of correcting the deficiency of TOPSIS and ELECTRE methods. Re- multi attribute decision-making problems. Computers and Industrial Engineering,
sults obtained from TOPSIS, ELECTRE and Grey Theory are pre- 55, 80–93.
sented in Table 10. The values in the table express the preference Lee, C. (1993). The multiproduct warehouse location problem: Applying a
decomposition algorithm. International Journal of Physical Distribution and
ranking of warehouse alternatives depending on the determined Logistics Management, 23, 3–13.
method. Li, G.D., Yamaguchi, D., & Nagai, M. (2006). A grey-based approach to suppliers
When TOPSIS and ELECTRE methods were used, similar results selection problem. In Proceedings 2006 of international conference on parallel and
distributed processing techniques and applications (PDPTA2006), Vol. 2 (pp. 818–
have been obtained despite the differences in calculation algo- 824).
rithms of decision methodologies. On the other hand, a warehouse Lopez, J. C., & Gonzalez, E. (2003). A new method for group decision support based
other than the results obtained with TOPSIS and ELECTRE has been on ELECTRE III methodology. European Journal of Operational Research, 148,
14–27.
selected as the best alternative. This difference is due to the differ- Olson, D. L. (2004). Compression of weights in TOPSIS models. Mathematical and
ences in formulas used at the standardization stage of performance Computer Modelling, 40, 721–727.
criteria of alternatives against criteria. Parkan, C., & Wu, M. L. (1995). Selection a manufacturing process with multiple
attributes: A case study. Journal of Engineering Technology Management, 12,
219–237.
References Poh, K. L., & Ang, B. W. (1999). Transportation fuels and policy for Singapore: An
AHP planning approach. Computers and Industrial Engineering, 37, 507–525.
Agrawal, V. P., Kohli, V., & Gupta, S. (1991). Computer aided robot selection: The Rogers, M., & Bruen, M. (1998). A new system for weighting environmental criteria
multiple attribute decision making approach. International Journal of Production for use within ELECTRE III. European Journal of Operational Research, 107,
Research, 29, 1629–1644. 552–563.
Anand, R. P. (1995). Multi-criteria methods in river basin planning: A case study. Roy, B. (1991). The outranking approach and the foundations of ELECTRE methods.
Water Science and Technology, 31, 261–272. Theory and Decision, 31, 49–73.
Chan, J. W. K., & Tong, T. K. L. (2007). Multi-criteria material selections and end-of- Saaty, T. L. (1980). The analytic hierarchy process. New-York: Mc-Graw Hill.
life product strategy: Grey relational analysis approach. Materials and Design, Shih, H., Shyur, H., & Lee, E. S. (2007). An extension of TOPSIS for group decision
28, 1539–1546. making. Mathematical and Computer Modelling, 45, 801–813.
Chau, O. L., & Parkan, C. (1995). Selection of a manufacturing process with multiple Tabari, M., Kaboli, A., Aryanezhad, M. B., Shahanaghi, K., & Siadat, A. (2008). A new
attributes: A case study. Production Planning and Control, 9, 189–200. method for location selection: A hybrid analysis. Applied Mathematics and
Chen, C. (2001). A fuzzy approach to select the location of the distribution center. Computation, 206, 598–606.
Fuzzy Sets and Systems, 118, 65–73. Wang, C. H. (2004). Predicting tourism demand using fuzzy time series and hybrid
Deng, J. L. (1982). Control problems of grey system. Systems and Control Letters, 1, Grey Theory. Tourism Management, 25, 367–374.
288–294. Wang, X., & Triantaphyllou, E. (2008). Ranking irregularities when evaluating
Deng, H., Yeh, C. H., & Willis, R. J. (2000). Inter-company comparison using modified alternatives by using some ELECTRE methods. The International Journal of
TOPSIS with objective weights. Computers& Operations Research, 10, 963–974. Management Science, 36, 45–63.
Feng, C. M., & Wang, R. T. (2001). Considering the financial ratios on performance Wen, K. (2004). The grey system analysis and its application in gas breakdown and
evaluation of highway bus industry. Transport Reviews, 21, 449–467. var compensator finding. International Journal of Computational Cognition, 2,
Figueira, J., & Roy, B. (2002). Determining the weights of criteria in the ELECTRE type 21–44.
methods with a revised Simos procedure. European Journal of Operational Yuan, X. (2007). Grey relational evaluation of financial situation of listed company.
Research, 139, 317–326. Journal of Modern Accounting and Auditing, 3, 41–44.

You might also like