Expert Systems with Applications 37 (2010) 4999–5008

Contents lists available at ScienceDirect

Expert Systems with Applications

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

A new TOPSIS-based multi-criteria approach to personnel selection

Alecos Kelemenis *, Dimitrios Askounis
National Technical University of Athens, Greece

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

Keywords: Selection of qualified human resources is a key success factor for an organization. The complexity and
Personnel selection importance of the problem call for analytical methods rather than intuitive decisions. The aim of this
Multi-criteria decision making paper is to support adequately the decision making process. The steps of fuzzy Technique for Order Pref-
Fuzzy TOPSIS erence by Similarity to an Ideal Solution (TOPSIS) are considered, incorporating a new concept for the
Veto threshold
ranking of the alternatives. This is based on the veto threshold, a critical characteristic of the main out-
ranking methods. The ultimate decision criterion is not the similarity to the ideal solution but the dis-
tance of the alternatives from the veto set by the decision makers. Additionally, a real life application
on the selection of a top management team member shows the practical implications.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction and growth (Bharadwaj, 2000; Bharadwaj, Sambamurthy, & Zmud,

1998; Mata et al., 1995).
It is a common belief that IT comprises a crucial factor for the Literature on IT management has reported more cases of failed
development and growth of an organization. Nevertheless, in prac- implementations than of success (Dhillon, 2008). Many communi-
tice, the IT function acts often as a support role in business rather cation and leadership inadequacies have been identified amongst
than as a leader and strategic partner (Ward & Peppard, 1996). This senior IT managers and consequent breakdowns in the IT/business
is true considering that scarcity makes a resource truly strategic, relationship (Willcoxson & Chatham, 2006). As Enns, Huff, and
while IT resources have become available and affordable to all Golden (2003) demonstrated, Chief Information Officers (CIOs)
(Carr, 2003). Studies recognize that many components of IT infra- usually lack the interpersonal and conceptual skills needed to
structure (such as off-the-shelf computer hardware and software) influence others in the organization. For instance, CIOs spend much
convey no particular strategic benefit due to lack of rarity, ease of their time attempting to convince other top managers to commit
of imitation, and ready mobility (Wade & Hulland, 2004). What to strategic IT initiatives (Lederer & Mendelow, 1988), share in a vi-
cannot be imitated are the managerial IT skills, in comparison to sion for IT (Earl & Feeny, 1994), and allocate resources to IT pro-
other IT resources. The study of Mata, Fuerst, and Barney (1995) jects (McKenney, Mason, & Copeland, 1997). Many IS/IT
empirically supported the link between managerial IT skills and professionals speak in jargon that shows a basic ignorance of the
firm performance. Powell and Dent-Micallef (1997) divided IT re- rest of the world of organizational leaders (Service, 2005). Taking
sources into three categories: human resources, business re- into account the above mentioned and the fact that technical qual-
sources, and technology resources. In a study of the US retail ifications should be considered as precondition, the critical dimen-
industry, they found that only human resources in concert with sions of human IT resources include: (a) technical IT skills, such as
IT contributed to improved performance. Firms with strong human programming, systems analysis and design, and competencies in
IT resources are able to (a) build internal relationships between the emerging technologies, and (b) the ‘‘soft” IT skills, which include
Information Systems (IS) function and other departments of the abilities such as information management skills, communication
firm, leading to integrated planning processes at corporate level, and negotiation skills, process and project management and lead-
(b) manage relationships between the IS function and stakeholders ership skills (Bharadwaj, 2000; Ward, 1999).
outside the firm, (c) anticipate future business needs of the firm Selection of IT professionals is then a critical factor for success-
and innovate valuable new product features before competitors ful IT management. In general, personnel selection, depending on
and in parallel manage effectively the resulting technology change the firm’s specific targets, the availability of means and the individ-
ual preferences of the decision makers (DMs), is a highly complex
* Corresponding author. Address: National Technical University of Athens, School problem. The multi-criteria nature of the problem makes Multi-
of Electrical and Computer Engineering, 9, Iroon Polytechniou str., Zografou 15773, Criteria Decision Making (MCDM) methods and fuzzy logic ideal
Athens, Greece. Tel.: +30 210 7723612/3583; fax: +30 210 7723550. to cope with this, given that they consider many criteria at the
E-mail address: [email protected] (A. Kelemenis).
same time, with various weights and thresholds, having the
URL: http://www.epu.ntua.gr/ (A. Kelemenis).

0957-4174/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.eswa.2009.12.013
5000 A. Kelemenis, D. Askounis / Expert Systems with Applications 37 (2010) 4999–5008

potential to reflect at a very satisfactory degree the vague – most of irrespectively of the values of the other criteria (Kangas, Kangas, &
the times – preferences of the DMs. Pykäläinen, 2001).
The rest of the paper is organized as follows: In the next section,
the main MCDM methods are summarized while some relevant 2.1.2. Multi-attribute utility and value theories
studies on the personnel selection problem are presented. In Sec- The multi-attribute utility theory (MAUT) tries to assign an
tion 3, the principles of the fuzzy sets are demonstrated in brief. overall utility value to each alternative. This utility is a real number
Section 4 presents the proposed approach to support the decision representing the preferability of the considered alternative.
making. Section 5 briefly presents an empirical application of the Weights that reflect the relative importance of the attributes are
proposed approach for the selection of a senior IT officer. Finally, defined. These weights are typically scaled so that they sum to
future steps and research challenges are discussed. unity. For each alternative a marginal utility value is assigned to
each attribute and the sum of the products of attribute marginal
utility and attribute importance is calculated. This weighted sum
2. Multi-criteria decision making methods
represents the overall utility value associated with the alternative.
The optimal alternative is the one with the higher overall utility
In most of the situations where a decision must be taken, it is
value (Figueira et al., 2005).
rare for the DM to have in mind a single clear criterion (Figueira,
The UTA (Utilités Additives) method was proposed by Jacquet-
Greco, & Ehrgott, 2005). Such situations, where a single-criterion
Lagreze and Siskos (1982). The UTA method refers to the philoso-
approach falls short, refer to as MCDM problems.
phy of assessing a set of value or utility functions, assuming the
Many terminologies have been proposed for the categorization
axiomatic basis of MAUT and adopting the preference disaggrega-
of MCDM problems. The dominant terms are the one of Multi-Cri-
tion principle. UTA methodology uses linear programming tech-
teria Decision Analysis (MCDA) or Multi-Attribute Decision Making
niques in order to optimally infer additive value/utility functions,
(MADM), for problems in which the DM must choose from a finite
so that these functions are as consistent as possible with the global
number of explicitly available alternatives characterized by a set of
DM’s preferences (inference principle). The behaviour and the cog-
multiple attributes (or criteria) and the one of Multi-Objective
nitive style of the DM are analyzed; special iterative interactive
Mathematical Programming (MOMP) or Multi-Objective Decision
procedures are used, where the components of the problem and
Making (MODM) that deal with decision problems characterized
the DM’s global judgment policy are analyzed and then they are
by multiple and conflicting objective functions that are to be opti-
aggregated into a value system.
mized over a feasible set of decisions. Here, the alternatives are not
The analytic hierarchy process (AHP), developed at the Wharton
explicitly known a priori (Figueira et al., 2005). In what follows, the
School of Business by Saaty (1980), allows the DMs to model a
main categories of MCDM are presented.
complex problem in a hierarchical structure showing the relation-
ships of the goal, objectives (criteria), sub-objectives, and alterna-
2.1. Multi-criteria decision analysis tives. AHP is based on three basic principles: decomposition,
comparative judgments, and hierarchic composition or synthesis
The aim in MCDA is to determine overall preferences among of priorities (Saaty, 1990). Decomposition principle is applied to
alternative options. In this context, the decision is facilitated by structure a complex problem into a hierarchy of clusters, sub-clus-
evaluating each option based on a set of criteria. The criteria must ters, sub-sub-clusters and so on. The principle of comparative judg-
be measurable and their outcomes must be measured for every ments is applied to construct pairwise comparisons of all
decision option. Criterion outcomes provide the basis for compar- combinations of elements in a cluster with respect to the parent
ison of option and consequently facilitate the final decision of of the cluster. These pairwise comparisons are used to derive
choice, sorting or ranking, depending on the specificities of the ‘‘local” priorities of the elements in a cluster with respect to their
decision problem. The MCDA methods can be divided into two parent. The principle of hierarchic composition or synthesis is ap-
main categories; the outranking methods (Roy, 1993) and the plied to multiply the local priorities of elements in a cluster by the
multi-attribute utility and value theories. ‘‘global” priority of the parent element, producing global priorities
throughout the hierarchy and then adding the global priorities for
2.1.1. Outranking methods the lowest level elements (the alternatives). Two of the main char-
Outranking methods are based on pairwise comparison of ac- acteristics of this method are that it enables DMs to derive ratio
tions. Outranking indicates the degree of dominance of one alter- scale priorities or weights as opposed to arbitrarily assigning them
native over another. The alternative a is deemed better than and that it is a compensatory decision methodology because alter-
alternative b if the number of criteria indicating that alternative natives that are deficient with respect to one or more objectives
a is better than alternative b is larger than the number of criteria can compensate by their performance with respect to other objec-
indicating the opposite. The usual case is the one of ranking the tives (Forman & Selly, 2001).
alternatives. The most commonly used outranking methods for The Technique for Order Preference by Similarity to Ideal Solu-
ranking problems are PROMETHE II and ELECTRE III. In these, crite- tion (TOPSIS) was firstly proposed by Hwang and Yoon (1981). The
ria are treated as so-called pseudo-criteria. This means that a approach is based on a synthesizing criterion like MAUT and AHP.
threshold model is applied to the original criteria value. If the cri- The main concept of this method is that the most preferred alter-
teria values are sufficiently close to each other, they are indifferent native should have the shortest distance from the positive ideal
to the DM (indifference threshold) and if the difference between solution (PIS) and the longest distance from the negative ideal
the criteria values is sufficiently large, there is no doubt which solution (NIS). PIS is the one that maximizes the benefit criteria
alternative is better according to that criterion (preference thresh- and minimizes the cost criteria, while the NIS maximizes the cost
old) (Figueira et al., 2005). In between, there is an area, in which criteria and minimizes the benefit criteria (Wang & Elhag, 2006).
the DM is assumed to hesitate between indifference and strict pref- In traditional TOPSIS, the weights of the criteria and the ratings
erence. In addition, these methods can be considered as non-com- of alternatives are known precisely and are treated as crisp numer-
pensatory models, meaning that a really bad score of any ical data. However, under many conditions crisp data are inade-
alternative with respect to any one criterion cannot necessarily quate to model real-life decision problems; in addition, perfect
be compensated for by good scores in other criteria. Thus, the alter- knowledge is not easily acquired. Unquantifiable, incomplete and
native with a very poor value of any one criterion cannot be chosen non-obtainable information (Ölçer & Odabasßi, 2005) make precise
 A. Kelemenis, D. Askounis / Expert Systems with Applications 37 (2010) 4999–5008 5001

judgement impossible. Therefore, fuzzy TOPSIS has been proposed from regular sets, also called crisp sets, to fuzzy sets we start with
where criteria weights and alternative ratings are given by linguis- crisp subsets of X. Let A be a subset of X. For each x in X we know
tic variables that are expressed by fuzzy numbers. whether x belongs or does not belong to A. Define a function on X
whose values are zero or one as follows: (a) the value of the func-
2.2. Multi-objective mathematical programming tion at x is one if x is a member of A; and (b) the value is zero if x
does not belong to A. We write this function as A(x) = 1 if x is in A
The other main category of MCDM problems is the one of and A(x) = 0 otherwise. This function is called the characteristic
MOMP. MOMP is a part of mathematical programming dealing function on A and any such function, whose values are either zero
with decision problems characterized by multiple and conflicting or one, defines a crisp subset of X.
objective functions that are to be optimized over a feasible set of Fuzzy sets generalize the characteristic function in allowing all
decisions. Such problems refer to as multi-objective programs values between zero and one. A fuzzy subset F of X is defined by its
(MOPs). Feasible alternatives are available implicitly, through con- membership function (a generalization of the characteristic func-
straints in the form of mathematical functions. An optimization tion), also written F(x), whose values can be any number in the
problem (typically a mathematical program) has to be solved to interval [0, 1]. The value of F(x) is called the grade of membership
explicitly find the alternatives. MOP may involve linear or nonlin- of x in fuzzy set F, and is often denoted by l(x). If l(x) is only zero
ear objective functions and constraints, and may have continuous or one, then we get the characteristic function of a crisp, non-fuzzy,
or integer decision variables (Figueira et al., 2005). Goal program- set F. Now suppose we have l(x) taking on values in [0, 1] besides
ming (GP), along with its many variants, is considered as one of the just zero and one. We say x belongs to F if lF(x) = 1, x does not be-
most effective strategies for solving MOMP (Abdelaziz, 2007). In GP long to F when lF(x) = 0, and x is in F with membership lF(x) if
one is interested in achieving a desirable goal or target established 0 < lF(x) < 1. The universal set always has lX(x) = 1 for all x in X,
for the objective functions of the MOP. and the empty set is described by its membership function always
zero [lO(x) = 0 for all x in X]. Crisp sets are considered special cases
of fuzzy sets when membership values are always 0 or 1 (Siler &
2.3. Personnel selection and MCDM
Buckley, 2005).
Among the MCDM problems that are encountered in real life is
the personnel selection problem. This problem, from the multi-cri- 3.2. Fuzzy numbers
teria perspective, has attracted the interest of many scholars. The
applications lie from internal auditor (Seol & Sarkis, 2005) and Fuzzy numbers represent a number of whose value we are
TQM consultant (Saremi, Mousavi, & Sanayei, 2009) to IS personnel somewhat uncertain. They are a special kind of fuzzy set whose
(Chen & Cheng, 2005) selection. members are numbers from the real line, and hence are infinite
Since it is a human resource management problem, the DMs in extent. Fuzzy numbers may be of almost any shape (though con-
face difficulties in assigning crisp values as scorings to the several ventionally they are required to be convex and to have finite area),
criteria. The fuzziness is the main characteristic of this problem but frequently they will be triangular (piecewise linear), s-shape
and most of the scholars extend the typical MCDM methods to (piecewise quadratic) or normal (bell shaped). Fuzzy numbers
the fuzzy environment (Chen & Cheng, 2005). In particular, there may also be basically trapezoidal, with an interval within which
is a significant volume of studies that extend AHP and TOPSIS in or- the membership is 1; such numbers are called fuzzy intervals. Fuz-
der to represent the scorings of the alternatives in fuzzy terms zy intervals may have linear, s-shape or normal ‘‘tails”, the increas-
(Chen, 2000; Güngöra, Serhadlioğlub, & Kesen, 2009; Saremi ing and decreasing slopes (Siler & Buckley, 2005). Assume that
et al., 2009). Another main feature that appears in the relevant lit- triangular and s-shaped fuzzy numbers start rising from zero at
erature is the one of the group decision making. Group decision x = l; reach a maximum of one at x= m; and decline to zero at
making approach has been proposed by many researchers (Chen, x = n. Then the membership function l(x) of a triangular fuzzy
2000; Chen & Cheng, 2005; Canós & Liern, 2008; Saremi et al., number is given by
2009; Shih, Shyur, & Lee, 2007). Moreover, aggregating the opin-
ions of group members is an issue of attention. A number of new 8
> 0; x<l
aggregation techniques have been presented, applied to the per- >
>
< xl ; l6x<m
ml
sonnel selection (Canós & Liern, 2008). Finally, some noteworthy lðxÞ ¼ ð1Þ
studies have come from the Artificial Intelligence field, proposing
>
>
xn
; m6x6n
>
: mn
data mining (Chien & Chen, 2008) and decision rules (Jereb, Rajko- 0; x>n
vic, & Rajkovic, 2005) techniques to support the selection of the
appropriate candidate. An example of a membership function is shown in Fig. 1. Member-
ship of 4 is 0.5.
3. Fuzzy logic

Most real world decision problems take place in a complex envi-

ronment where conflicting systems of logic, uncertain and imprecise
knowledge, and possibly vague preferences have to be considered.
To face such complexity, the use of specific tools, techniques, and
concepts which allow the available information to be represented
with the appropriate granularity is believed as crucial. Particularly,
fuzzy set theory can ideally cope with this kind of problems.

3.1. Fuzzy sets

Let X be a collection of objects called a universal set. Every other

collection of objects will be subset of X. To explain the transition Fig. 1. Membership function of a triangular fuzzy number.
5002 A. Kelemenis, D. Askounis / Expert Systems with Applications 37 (2010) 4999–5008

4. Proposed approach form the group of the DMs. Nevertheless, there are situations in
which some criteria are too technical to allow each DM to express
In this paper, the process of fuzzy TOPSIS is considered, incorpo- a comprehensive opinion. Against these criteria, only few experts
rating a new measurement for the ranking of the alternatives, can provide input and judge objectively. Thus, group decision mak-
based on the veto concept, a critical characteristic of the main out- ing may not be the appropriate method.
ranking methods. The ultimate decision criterion is not the similar-
ity to the ideal solution but the distance from the imposed vetos 4.2. Definition of a finite set of relevant criteria
thresholds, as they will be defined in this section.
As mentioned in Section 2.3, TOPSIS has been used in a number Criteria should be defined that cover the requirements of the
of personnel selection problems. This is one of the many applica- DMs and relate to the specific job description. The process should
tions of the method, appeared in the literature. These applications take into consideration the market, in which the firm operates, the
lie among the several aspects of classical business, such as market- type and the hierarchical level of the position to be covered. More-
ing, supply chain, manufacturing as well as the high-tech contem- over, criteria associated to the required employee profile (e.g., per-
porary problems. Some indicative recent studies are the following: sonality traits and the so-called social skills) should be embedded
In manufacturing sector, Wang, Cheng, and Huang (2009) Chen, in the model. Thus, apart from a number of generic criteria, specific
Lin, and Huang (2006) dealt with the supplier selection problem. ones should be considered case by case. For example, different cri-
The latter proposed a simplified parameterized metric distance to teria should be considered for salesmen, high level managers,
calculate the distance between each point and fuzzy PIS and fuzzy developers or senior accountants.
NIS. Perçin (2008) presented the employment of a new hierarchical
fuzzy TOPSIS approach to evaluate the most suitable business pro- 4.3. Choice of appropriate linguistic variables and respective scales for
cess outsourcing (BPO) decision and Bottani and Rizzi (2006) pre- the weights of the criteria and the ratings of the alternatives
sented a TOPSIS-based approach for the selection and ranking of
the most suitable 3PL service provider, while Shyjith, Ilangkuma- A linguistic variable is a variable whose values are linguistic
ran, and Kumanan (2008) focused on the use of AHP and TOPSIS terms, i.e. words or sentences (Siler & Buckley, 2005). For example,
to select an optimum maintenance strategy for a textile industry. communication skill is a linguistic variable when its linguistic val-
Since high-technology has entered the manufacturing industry, ues are poor, fair, good. Each linguistic value can be represented by
specialized industrial robotic systems selection were examined in a fuzzy number which can be assigned to a membership function.
Kahraman, Çevik, Ates, and Gülbay (2007), incorporating technical In our approach, we consider triangular fuzzy numbers to be asso-
and economical factors in their evaluation through TOPSIS method. ciated to the linguistic variables.
In the same line, Kahraman, Ates, Cevik, Gulbay, and Erdogan A different decision on the scale used could lead to different re-
(2007) developed a multi-attribute decision making model for sults, a fact that cannot ensure the ‘‘objectivity” of the overall deci-
evaluating and selecting among logistic information technologies sion process. In this work, considering previous studies, scales of
using hierarchical TOPSIS. Marketing application can be found in seven points for the weights of the criteria and the ratings of the
the work of Kahraman, Büyüközkan, and Atesß (2007), supporting alternatives are suggested.
the development of a new product. Researchers have turned their
attention also to tourism management, in an attempt to facilitate 4.4. Selection of criteria weights for each DM
tourists’ choice of destination (Hsu, Tsai, & Wu, 2009) and evaluate
the service quality of a number of hotels (Benítez, Martín, & Romá- Each DM should assign the importance of each criterion, in lin-
n, 2007). A synthesis of Fuzzy AHP and TOPSIS methods was used guistic terms, reflecting to a specific fuzzy number, according to
in Ertuğrul and Karakasßoğlu (2009), evaluating firms’ performance, the requirements and the expectations from the position to be
while a comparison between these two multi-criteria decision sup- filled. In group decision making cases, it is not uncommon that
port methods for the facility location selection was used by the two DMs have conflicting views on the importance of a criterion,
same researchers (Ertuğrul & Karakasßoğlu, 2008). one considering it very important while the other does not. A great
In general, TOPSIS method is easy to understand and to imple- volume of research has focused on weighting methods; however
ment. These issues are of fundamental importance for a direct field this is beyond the scopes of this work.
implementation of the methodology by practitioners. Moreover, it
allows the straight linguistic definition of weights and ratings un- 4.5. Determination of the veto threshold for each DM
der each criterion, without the need of cumbersome pairwise com-
parisons and the risk of inconsistencies (Bottani & Rizzi, 2006). In order to simulate the reality and the behaviour of the DMs,
Also, according to Zanakis, Solomon, Wishart, and Dublish veto thresholds should be defined by every DM.
(1998), the performance is slightly affected by the number of alter- In outranking methods, veto threshold indicates situations
natives and rank discrepancies are amplified to a lesser extent for when the difference between two alternatives with respect to
increasing values of the number of alternatives and the number of one specified criterion negates any possible outranking relation-
criteria. ship indicated by other criteria. This means that when an alterna-
In order to incorporate the specific preferences of the DMs, we tive is significantly bad on one criterion compared to another
integrate the fuzzy TOPSIS using the veto concept. alternative, it cannot outrank the latter, regardless its performance
The steps of the proposed method can be described as follows. on the other criteria.
We ‘‘borrow” this concept, allowing each DM to assign a veto to
each criterion. In this respect, veto expresses the power of every
4.1. Formation of the decision making group DM to negate the selection of an alternative as a solution, when
this alternative performs worse than the veto set on the respective
Personnel selection is a critical task for an organization, in par- criterion.
ticular when it is a part of a broader process at corporate level (e.g., Since we may face situations when all alternatives perform be-
strategic change). In this process, more rational decisions are made low a veto, we propose that the preferred alternative is the one
by a group of people rather than by a single person. Usually, ex- with the higher (positive) aggregated distance from the vetos of
perts from different departments along with high level managers all criteria.
 A. Kelemenis, D. Askounis / Expert Systems with Applications 37 (2010) 4999–5008 5003

4.6. Construction of the fuzzy decision matrix this respect, we can cross check the results and obtain the most
representative ranking.
The fuzzy decision matrix is presented as follows: Suppose two triangular fuzzy numbers u = (l, m, n) and
2 3 w = (a, b, c) and the original point (0, 0, 0).
~x11 ~x12  ~x1n
The membership function of the fuzzy number u = (l, m, n) can
6 ~x x~22    ~x2n 7
6 21 7 be expressed as
~ ¼6
D 6 .. .. .. .. 7
7 ð2Þ 8
4 . . . . 5 > 0; x<l
>
> L
~xm1 ~xm2    ~xmn <
l ðxÞ; l 6 x < m
f ¼ ½w
lðxÞ ¼ ð13Þ
>
> l ðxÞ; m 6 x 6 n
R
W ~ 1; w ~ 2; . . . ; w ~ n ð3Þ >
:
Te ¼ ½~t 1 ; ~t 2 ; . . . ; ~t n  ð4Þ 0; x>n
1 1 and l the inverse function of l. The methods to be used are:
~xij ¼ ½x ðþÞx2ij ðþÞ . . . ðþÞxKij  ð5Þ
K ij
1 1 (a) Sign distance, proposed by Abbasbandy and Asady (2006).The
~ j ¼ ½w
w ~ ðþÞw ~ 2j ðþÞ . . . ðþÞw ~ Kj  ð6Þ
K j distance from the original point is
~t j ¼ 1 ½~t1 ðþÞ~t 2 ðþÞ . . . ðþÞ~t K  ð7Þ dðu; 0Þ ¼ cðuÞ  Dðu; 0Þ ð14Þ
K j j j

~ j ; ~t j ; i ¼ 1; 2; . . . ; m; j ¼ 1; 2; . . . ; n are linguistic with c(u) = 1 if inf u(r) P 0 and c(u) = 1 if sup u

 ðrÞ < 0 and
where ~xij ; 8i; j; w
variables that are described by triangular fuzzy numbers uðrÞ ¼ l þ ðm  lÞ  r ð15Þ
~ j ¼ ðwj1 ; wj2 ; wj3 Þ and ~t j ¼ ðt j1 ; tj2 ; t j3 Þ.
~xij ¼ ðaij ; bij ; cij Þ; w  ðrÞ ¼ n  ðn  mÞ  r
u ð16Þ
K is the number of DMs, m is the number of alternatives, n is the
number of criteria, W f is the matrix of the weights and T e is the ma- In our case, it is always c(u) = 1, due to the meanings of the
trix of the veto thresholds. fuzzy numbers (i.e., scorings of the criteria and the alterna-
tives). D(u, 0) is calculated from the formula (17)
4.7. Construction of the normalized fuzzy matrix Z 1   1P
P
Dðu; 0Þ ¼  ðrÞjP dr
juðrÞj þ ju P P 1; ð17Þ
Applying the appropriate operator to each element of the fuzzy 0

decision matrix, we obtain the fuzzy normalized matrix. In this re- So, the distance between two triangular fuzzy numbers, u and
spect, every triangular fuzzy number lies between zero and one. w is
The matrix can be expresses as
dðu; wÞ ¼ dðu; 0Þ  dðw; 0Þ ð18Þ
~ ¼ ½~rij 
R ð8Þ
mn
! In this study, P = 1 and P = 2 will be used.
aij bij cij (b) Signed distance, proposed by Yao and Wu (2000).The distance
~r ij ¼ ; ; ; j2B ð9Þ
cj cj cj from the original point is
cj ¼ max cij if j 2 B Z
i
1 1
dðu; 0Þ ¼ ½l þ ðm  lÞ  r þ n  ðn  mÞ  r dr
2 0
where B is the set of benefit criteria and
1
    ¼ ð2  m þ l þ nÞ ð19Þ
aj aj aj 4
~rij ¼ ; ; ; j2C
cij bij aij ð10Þ Again, the distance between two triangular fuzzy numbers, u
aj ¼ min aij if j 2 C and w is given by (18).
i
(c) Cheng’s (1998) method.This is based on the distance index
where C is the set of cost criteria, following the linear normalization between the centroid point ð Þ as defined in formulas (20)
x; y
method (Shih et al., 2007). and (21)
Rm L Rn
xl ðxÞdx þ m xlR ðxÞdx
4.8. Construction of the weighted normalized fuzzy matrix xðuÞ ¼ lR m R ð20Þ
l
lL ðxÞdx þ mn lR ðxÞdx
R 1 L R1
In this step, we incorporate the importance of each criterion, yl ðyÞdy þ 0 ylR ðyÞdy
ðuÞ ¼ 0R 1
y R ð21Þ
taking the matrix lL ðyÞdy þ 01 lR ðyÞdy
0
 
V~ ¼ v~ ij mn ; i ¼ 1; 2; . . . m; j ¼ 1; 2; . . . n ð11Þ and the original point, i.e.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where
RðuÞ ¼ ðxÞ2 þ ðy  Þ2 ð22Þ
v~ ij ¼ ~rij ðÞw~ j ð12Þ
The distance between two triangular fuzzy numbers, u and w
is
4.9. Calculation of the distance of each alternative from the veto
dðu; wÞ ¼ RðuÞ  RðwÞ ð23Þ
threshold defined for each criterion
(d) Chu and Tsao’s (2002) method.This is close to the concept of
This is the last step before the ranking of the alternatives, in Cheng’s (1998) study, but the distance is calculated by
which we must compare each alternative with the veto threshold
SðuÞ ¼ xðuÞ  y
ðuÞ ð24Þ
defined for each criterion.
In order to calculate the distance, we propose the use of six dif- instead of R(u) and the distance between two triangular fuzzy
ferent approaches presented in the literature, simultaneously. In numbers, u and w is
5004 A. Kelemenis, D. Askounis / Expert Systems with Applications 37 (2010) 4999–5008

dðu; wÞ ¼ SðuÞ  SðwÞ ð25Þ Table 2

Scale for scoring the alternatives.
(e) Yager’s (1980) method. Using the 
x value on the horizontal
Ratings Fuzzy sets
axis (20) as the only index for comparison, the distance
between two triangular fuzzy numbers, u and w is Very poor (VP) (0, 0, 1)
Poor (P) (0, 1, 3)
dðu; wÞ ¼ xðuÞ  xðwÞ ð26Þ Medium poor (MP) (1, 3, 5)
Fair (F) (3, 5, 7)
Medium good (MG) (5, 7, 9)
Finally, the distance of each alternative from the veto of each Good (G) (7, 9, 10)
criterion j, j = 1, 2, . . . , n is Very good (VG) (9, 10, 10)

V
X
n
di ¼ dðv~ ij ; ~t j Þ; i ¼ 1; 2; . . . ; m ð27Þ
j¼1 Table 3
‘‘Soft” skills criteria weights.

Criteria DM1
4.10. Rank the alternatives
C1 VH
C2 VH
The preferred alternative is the one with the higher (positive) C3 H
distance from the vetos of all criteria. Thus, the one with the max- C4 H
V
imum value of di is the best alternative. C5 M

5. An empirical application
Table 4
The purpose of the empirical application was to illustrate the Technical skills criteria weights.
use of the suggested method. The experiment was basically setup
Criteria DM2
upon a real life decision. A branch office of a multinational IT firm
C6 H
A, wanted to recruit a CIO externally, since the previous one moved
C7 VH
to the Headquarters. A group of consultants to the Chief Executive C8 VH
Officer (CEO) made a final choice of four candidates in order that C9 H
the CEO selects one among them. After a meeting with the CEO, C10 M
we mutually agree that one external expert in organizational psy- C11 MH

chology field would be the evaluator of the ‘‘soft” skills required for
the position of the CIO, while the CEO, being an IT expert himself,
would score the various technical skills.
All involved actors resulted in a set of 11 criteria, i.e. strategy Table 5
‘‘Soft” skills criteria veto thresholds.
formulation/strategic decision making (C1), change management/
change adaptability (C2), communication/interpersonal skill (C3), Criteria DM1
leadership (C4), risk/crisis management (C5), computer networks C1 G
(C6), software/software tools (C7), databases (C8), professional C2 G
experience (C9), educational background (C10), emerging/new C3 MP
C4 MP
technologies (C11).
C5 P
According to the methodology described in the previous sec-
tion, scales of seven points for the weights of the criteria and the
ratings of the alternatives were considered, expressed in triangular
fuzzy numbers, as shown in the Tables 1 and 2.
Table 6
The external consultant (DM1) was asked to define the impor- Technical skills criteria veto thresholds.
tance of each ‘‘soft” criterion, according to his preferences and pri-
Criteria DM2
orities. Table 3 depicts his preferences, expressed in linguistic
variables. C6 MG
C7 G
Accordingly, the CEO (DM2) was asked to define the importance
C8 MG
of each technical criterion (Table 4). C9 MG
Then, the DMs had also to define the veto threshold for each cri- C10 F
terion; this means the lower ‘‘allowed” performance of each alter- C11 MP
native to each criterion in order that an alternative was not
considered as inappropriate (Tables 5 and 6).
Then, scores for each alternative in each criterion were assigned
Table 1 by the DMs (Table 7).
Scale for weighting the criteria. The fuzzy decision matrix was then constructed following the
Weights Fuzzy sets
scores and the preferences of the DMs (Table 8).
Table 9 depicts the normalized fuzzy decision matrix.
Very low (VL) (0, 0, 0.1)
Low (L) (0, 0.1, 0.3)
Table 10 shows the weighted normalized fuzzy decision matri-
Medium low (ML) (0.1, 0.3, 0.5) ces, taking into consideration the importance of each criterion.
Medium (M) (0.3, 0.5, 0.7) The comparison among the different distance measurement is
Medium high (MH) (0.5, 0.7, 0.9) shown in Table 11.
High (H) (0.7, 0.9, 1)
Five out of six methods produced exactly the same ranking,
Very high (VH) (0.9, 1, 1)
while all of them resulted in the same alternative as the best
 A. Kelemenis, D. Askounis / Expert Systems with Applications 37 (2010) 4999–5008 5005

Table 7 Table 8
Scores of the alternatives in criteria (linguistic values). Fuzzy decision matrix.

Criteria Alternatives Scores (DM1/DM2) Criteria Alternatives Scores

C1 A1 VG C1 A1 (9, 10, 10)
A2 F A2 (3, 5, 7)
A3 VG A3 (9, 10, 10)
A4 G A4 (7, 9, 10)
Weight (0.9, 1, 1)
C2 A1 F
Veto (7, 9, 10)
A2 G
A3 G C2 A1 (3, 5, 7)
A4 G A2 (7, 9, 10)
A3 (7, 9, 10)
C3 A1 G
A4 (7, 9, 10)
A2 VG
Weight (0.9, 1, 1)
A3 P
Veto (7, 9, 10)
A4 F
C3 A1 (7, 9, 10)
C4 A1 F
A2 (9, 10, 10)
A2 VG
A3 (0, 1, 3)
A3 F
A4 (3, 5, 7)
A4 F
Weight (0.7, 0.9, 1)
C5 A1 G Veto (1, 3, 5)
A2 F
C4 A1 (3, 5, 7)
A3 VG
A2 (9, 10, 10)
A4 VG
A3 (3, 5, 7)
C6 A1 G A4 (3, 5, 7)
A2 VG Weight (0.7, 0.9, 1)
A3 P Veto (1, 3, 5)
A4 VG
C5 A1 (7, 9, 10)
C7 A1 G A2 (3, 5, 7)
A2 F A3 (9, 10, 10)
A3 VG A4 (9, 10, 10)
A4 G Weight (0.3, 0.5, 0.7)
C8 A1 G Veto (0, 1, 3)
A2 VG C6 A1 (7, 9, 10)
A3 F A2 (9, 10, 10)
A4 G A3 (0, 1, 3)
C9 A1 G A4 (9, 10, 10)
A2 G Weight (0.7, 0.9, 1)
A3 F Veto (5, 7, 9)
A4 F C7 A1 (7, 9, 10)
C10 A1 G A2 (3, 5, 7)
A2 G A3 (9, 10, 10)
A3 VG A4 (7, 9, 10)
A4 F Weight (0.9, 1, 1)
Veto (7, 9, 10)
C11 A1 G
A2 F C8 A1 (7, 9, 10)
A3 VG A2 (9, 10, 10)
A4 VG A3 (3, 5, 7)
A4 (7, 9, 10)
Weight (0.9, 1, 1)
Veto (5, 7, 9)
one. Thus, according to the distance of the alternatives from the C9 A1 (7, 9, 10)
veto thresholds set by the DMs, A1 is the one to select. A2 (7, 9, 10)
The final ranking obtained by the proposed approach was to- A3 (3, 5, 7)
A4 (3, 5, 7)
tally in accordance with the intuitive preference of the DMs. This Weight (0.7, 0.9, 1)
is an indication of the usefulness of this approach. Veto (5, 7, 9)
C10 A1 (7, 9, 10)
A2 (7, 9, 10)
6. Discussion and future research challenges A3 (9, 10, 10)
A4 (3, 5, 7)
The aim of this paper was to support adequately the decision on Weight (0.3, 0.5, 0.7)
IT professional selection. Even if a few years ago this was not the Veto (3, 5, 7)
case, undoubted great effect of IT to firm’s efficiency has ‘‘forced” C11 A1 (7, 9, 10)
CEOs to consider IT department a strategic partner as regards the A2 (3, 5, 7)
A3 (9, 10, 10)
long term goals and the direction to follow towards their achieve-
A4 (9, 10, 10)
ments. CIO is now a member of the top management team and Weight (0.5, 0.7, 0.9)
takes part in the decision making at corporate level. In general, Veto (1, 3, 5)
IT professionals must have the hard and soft skills to deal with
the rapidly changeful and uncertain business settings, in which
technology and knowledge are easily scorned while human rela-
tions in the workplace are becoming more and more personalized Selecting the appropriate IT professionals is a key success factor
and are related to firm’s performance at a wide extent. for an organization. The complexity and importance of the problem
5006 A. Kelemenis, D. Askounis / Expert Systems with Applications 37 (2010) 4999–5008

Table 9 Table 10
Normalized fuzzy decision matrix. Weighted normalized fuzzy decision matrix.

Criteria Alternatives Scores Criteria Alternatives Scores

C1 A1 (0.9, 1, 1) C1 A1 (0.81, 1, 1)
A2 (0.3, 0.5, 0.7) A2 (0.27, 0.5, 0.7)
A3 (0.9, 1, 1) A3 (0.81, 1, 1)
A4 (0.7, 0.9, 1) A4 (0.63, 0.9, 1)
Veto (0.7, 0.9, 1) Veto (0.63, 0.9, 1)
C2 A1 (0.3, 0.5, 0.7) C2 A1 (0.27, 0.5, 0.7)
A2 (0.7, 0.9, 1) A2 (0.63, 0.9, 1)
A3 (0.7, 0.9, 1) A3 (0.63, 0.9, 1)
A4 (0.7, 0.9, 1) A4 (0.63, 0.9, 1)
Veto (0.7, 0.9, 1) Veto (0.63, 0.9, 1)
C3 A1 (0.7, 0.9, 1) C3 A1 (0.49, 0.81, 1)
A2 (0.9, 1, 1) A2 (0.63, 0.9, 1)
A3 (0, 0.1, 0.3) A3 (0, 0.09, 0.3)
A4 (0.3, 0.5, 0.7) A4 (0.21, 0.45, 0.7)
Veto (0.1, 0.3, 0.5) Veto (0.07, 0.27, 0.5)
C4 A1 (0.3, 0.5, 0.7) C4 A1 (0.21, 0.45, 0.7)
A2 (0.9, 1, 1) A2 (0.63, 0.9, 1)
A3 (0.3, 0.5, 0.7) A3 (0.21, 0.45, 0.7)
A4 (0.3, 0.5, 0.7) A4 (0.21, 0.45, 0.7)
Veto (0.1, 0.3, 0.5) Veto (0.07, 0.27, 0.5)
C5 A1 (0.7, 0.9, 1) C5 A1 (0.21, 0.45, 0.7)
A2 (0.3, 0.5, 0.7) A2 (0.09, 0.25, 0.49)
A3 (0.9, 1, 1) A3 (0.27, 0.5, 0.7)
A4 (0.9, 1, 1) A4 (0.27, 0.5, 0.7)
Veto (0, 0.1, 0.3) Veto (0, 0.05, 0.21)
C6 A1 (0.7, 0.9, 1) C6 A1 (0.49, 0.81, 1)
A2 (0.9, 1, 1) A2 (0.63, 0.9, 1)
A3 (0, 0.1, 0.3) A3 (0, 0.09, 0.3)
A4 (0.9, 1, 1) A4 (0.63, 0.9, 1)
Veto (0.5, 0.7, 0.9) Veto (0.35, 0.63, 0.9)
C7 A1 (0.7, 0.9, 1) C7 A1 (0.63, 0.9, 1)
A2 (0.3, 0.5, 0.7) A2 (0.27, 0.5, 0.7)
A3 (0.9, 1, 1) A3 (0.81, 1, 1)
A4 (0.7, 0.9, 1) A4 (0.63, 0.9, 1)
Veto (0.7, 0.9, 1) Veto (0.63, 0.9, 1)
C8 A1 (0.7, 0.9, 1) C8 A1 (0.63, 0.9, 1)
A2 (0.9, 1, 1) A2 (0.81, 1, 1)
A3 (0.3, 0.5, 0.7) A3 (0.27, 0.5, 0.7)
A4 (0.7, 0.9, 1) A4 (0.63, 0.9, 1)
Veto (0.5, 0.7, 0.9) Veto (0.45, 0.7, 0.9)
C9 A1 (0.7, 0.9, 1) C9 A1 (0.49, 0.81, 1)
A2 (0.7, 0.9, 1) A2 (0.49, 0.81, 1)
A3 (0.3, 0.5, 0.7) A3 (0.21, 0.45, 0.7)
A4 (0.3, 0.5, 0.7) A4 (0.21, 0.45, 0.7)
Veto (0.5, 0.7, 0.9) Veto (0.35, 0.63, 0.9)
C10 A1 (0.7, 0.9, 1) C10 A1 (0.21, 0.45, 0.7)
A2 (0.7, 0.9, 1) A2 (0.21, 0.45, 0.7)
A3 (0.9, 1, 1) A3 (0.27, 0.5, 0.7)
A4 (0.3, 0.5, 0.7) A4 (0.09, 0.25, 0.49)
Veto (0.3, 0.5, 0.7) Veto (0.09, 0.25, 0.49)
C11 A1 (0.7, 0.9, 1) C11 A1 (0.35, 0.63, 0.9)
A2 (0.3, 0.5, 0.7) A2 (0.15, 0.35, 0.63)
A3 (0.9, 1, 1) A3 (0.45, 0.7, 0.9)
A4 (0.9, 1, 1) A4 (0.45, 0.7, 0.9)
Veto (0.1, 0.3, 0.5) Veto (0.05, 0.21, 0.45)

call for analytical methods rather than intuitive decisions. Taking ation in our approach. Thus, a new approach based on fuzzy TOPSIS
into consideration the above mentioned, the aim of this study for was used, introducing a new measurement. This is the veto
was to deal with IT professional selection problem, proposing a threshold that reflects the minimum requirements of the DMs
multi-criteria decision analysis approach. from each alternative in each criterion.
The specificity of this problem consists in dealing with impre- The proposed approach has practical implications as the empir-
cise data, difficulties in retrieving information and expressing an ical test showed in the case of a CIO selection problem of a multi-
explicit opinion. Fuzzy logic is considered ideal to deal with this national firm, supporting a very sensitive decision in real time.
type of problems. Moreover, every decision problem is closely As a future step to this paper could be the comparison of the
associated to the DMs. Every DM has specific preferences and de- proposed approach to other MCDA methods, like TOPSIS, AHP or
mands prerequisites in relation to the profile of the ideal solution. even more to the outranking methods, such as ELECTRE III and
Both above mentioned characteristics were taken into consider- PROMETHEE II. Moreover, situations should be studied, in which
 A. Kelemenis, D. Askounis / Expert Systems with Applications 37 (2010) 4999–5008 5007

Table 11 Chen, C. T. (2000). Extensions of the TOPSIS for group decision-making under fuzzy
Comparative results of different distance measurements. environment. Fuzzy Sets and Systems, 114(1), 1–9.
Chen, C. T., Lin, C. T., & Huang, S. F. (2006). A fuzzy approach for supplier evaluation
Methods Alternatives Distance values Ranking and selection in supply chain management. International Journal of Production
Economics, 102(2), 289–301.
Yager
Chen, L. S., & Cheng, C. H. (2005). Selecting IS personnel use fuzzy GDSS based on
A1 1.773 1
metric distance method. European Journal of Operational Research, 160(3),
A2 1.537 2 803–820.
A3 0.410 4 Cheng, C. H. (1998). A new approach for ranking fuzzy numbers by distance method.
A4 1.400 3 Fuzzy Sets and Systems, 95(3), 307–317.
Sign distance (Abbasbandy and Asandy, P = 1) Chien, C. F., & Chen, L. F. (2008). Data mining to improve personnel selection and
A1 3.660 1 enhance human capital: A case study in high-technology industry. Expert
Systems with Applications, 34(1), 280–290.
A2 3.180 2
Chu, T. C., & Tsao, C. T. (2002). Ranking fuzzy numbers with an area between the
A3 0.850 4
centroid point and original point. Computers and Mathematics with Applications,
A4 2.895 3
43(1), 111–117.
Sign distance (Abbasbandy and Asandy, P = 2) Dhillon, G. (2008). Organizational competence for harnessing IT: A case study.
A1 2.529 1 Information and Management, 45(5), 297–303.
A2 2.172 2 Earl, M. J., & Feeny, D. F. (1994). Is your CIO adding value. Sloan Management Review,
A3 0.560 4 35(3), 11–20.
Enns, H. G., Huff, S. L., & Golden, B. R. (2003). CIO influence behaviors: The impact of
A4 1.979 3
technical background. Information and Management, 40(5), 467–485.
Signed distance (Yao and Wu) _ & Karakasßoğlu, N. (2008). Comparison of fuzzy AHP and fuzzy TOPSIS
Ertuğrul, I.,
A1 1.648 1 methods for facility location selection. International Journal of Advanced
A2 1.408 3 Manufacturing Technology, 39(7/8), 783–795.
_ & Karakasßoğlu, N. (2009). Performance evaluation of Turkish cement
Ertuğrul, I.,
A3 0.203 4
A4 1.448 2 firms with fuzzy analytic hierarchy process and TOPSIS methods. Expert Systems
with Applications, 36(1), 702–715.
Cheng Figueira, J., Greco, S., & Ehrgott, M. (2005). Multiple criteria decision analysis: State of
A1 1.224 1 the art surveys series. Boston, MA: Springer.
A2 1.086 2 Forman, E. H., & Selly, M. A. (2001). Decision by objectives: How to convince others
A3 0.276 4 that you are right. New Jersey, NJ: World Scientific Publishing.
A4 0.962 3 Güngöra, Z., Serhadlioğlub, G., & Kesen, S. E. (2009). A fuzzy AHP approach to
personnel selection problem. Applied Soft Computing, 9(2), 641–646.
Chu and Tsao Hsu, T. K., Tsai, Y. F., & Wu, H. H. (2009). The preference analysis for tourist choice of
A1 0.916 1 destination: A case study of Taiwan. Tourism Management, 30(2), 288–297.
A2 0.795 2 Hwang, C. L., & Yoon, K. (1981). Multiple attributes decision making methods and
A3 0.212 4 applications. Berlin: Springer.
A4 0.724 3 Jacquet-Lagreze, E., & Siskos, J. (1982). Assessing a set of additive utility functions
for multicriteria decision making: The UTA method. European Journal of
Operational Research, 10(2), 151–164.
Jereb, E., Rajkovic, U., & Rajkovic, V. (2005). A hierarchical multi-attribute system
a group of experts, each one of different importance, are involved approach to personnel selection. International Journal of Selection and
in the decision making process. In this situation, the aggregation Assessment, 13(3), 198–205.
process or the consensus among the DMs is a main issue of Kahraman, C., Ates, N. Y., Cevik, S., Gulbay, M., & Erdogan, S. A. (2007). Hierarchical
fuzzy TOPSIS model for selection among logistics information technologies.
concern. Journal of Enterprise Information Management, 20(2), 143–168.
Kahraman, C., Büyüközkan, G., & Atesß, N. Y. (2007). A two phase multi-attribute
decision-making approach for new product introduction. Information Sciences,
Acknowledgment 177(7), 1567–1582.
Kahraman, C., Çevik, S., Ates, N. Y., & Gülbay, M. (2007). Fuzzy multi-criteria
evaluation of industrial robotic systems. Computers and Industrial Engineering,
This work has been funded by the project PENED 2003. The pro- 52(4), 414–433.
ject is co-financed 80% of public expenditure through EC – Euro- Kangas, A., Kangas, J., & Pykäläinen, J. (2001). Outranking methods as tools in
pean Social Fund, 20% of public expenditure through Ministry of strategic natural resources planning. Silva Fennica, 35(2), 215–227.
Lederer, A. L., & Mendelow, A. L. (1988). Convincing top management of the
Development – General Secretariat of Research and Technology strategic potential of information systems. MIS Quarterly, 12(4), 525–534.
and through private sector, under measure 8.3 of Operational Pro- Mata, F. J., Fuerst, W. L., & Barney, J. B. (1995). Information technology and sustained
gramme ‘‘Competitiveness” in the 3rd Community Support competitive advantage: A resource-based analysis. MIS Quarterly, 19(4), 487–505.
McKenney, J. L., Mason, R. O., & Copeland, D. G. (1997). Bank of America: The crest
Programme. and trough of technological leadership. MIS Quarterly, 21(3), 321–353.
_ & Odabasßi, A. Y. (2005). A new fuzzy multiple attributive group decision
Ölçer, A. I.,
making methodology and its application to population/maneuvering system
References selection problem. European Journal of Operational Research, 166(1), 93–114.
Perçin, S. (2008). Fuzzy multi-criteria risk-benefit analysis of business process
outsourcing (BPO). Information Management and Computer Security, 16(3),
Abbasbandy, S., & Asady, B. (2006). Ranking of fuzzy numbers by sign distance.
213–234.
Information Sciences, 176(16), 2405–2416.
Powell, T. C., & Dent-Micallef, A. (1997). Information technology as competitive
Abdelaziz, F. B. (2007). Multiple objective programming and goal programming:
advantage: The role of human, business, and technology resources. Strategic
New trends and applications. European Journal of Operational Research, 177(3),
Management Journal, 18(5), 375–405.
1520–1522.
Roy, B. (1993). Aide multicritère à la décision, méthodes et cas. Paris: Economica.
Benítez, J. M., Martín, J. C., & Román, C. (2007). Using fuzzy number for measuring
Saaty, T. L. (1980). The analytic hierarchy process. New York, NY: McGraw-Hill.
quality of service in the hotel industry. Tourism Management, 28(2), 544–555.
Saaty, T. L. (1990). Multicriteria decision making: The analytic hierarchy process:
Bharadwaj, A. S. (2000). A resource-based perspective on information technology
Planning, priority setting, resource allocation. Pittsburgh, PA: RWS Publications.
capability and firm performance: An empirical investigation. MIS Quarterly,
Saremi, M., Mousavi, S. F., & Sanayei, A. (2009). TQM consultant selection in SMEs
24(1), 169–196.
with TOPSIS under fuzzy environment. Expert Systems with Applications, 36(2),
Bharadwaj, A. S., Sambamurthy, V., & Zmud, R. W. (1998). IT capabilities:
2742–2749.
Theoretical perspectives and empirical operationalization. In R. Hirschheim,
Seol, I., & Sarkis, J. (2005). A multi-attribute model for internal auditor selection.
M. Newman, J. I. DeGross (Eds.), Proceedings of the 19th international conference
Managerial Auditing Journal, 20(8), 876–892.
on information systems, Helsinki.
Service, R. W. (2005). CQ: The communication quotient for IS professionals. Journal
Bottani, E., & Rizzi, A. (2006). A fuzzy TOPSIS methodology to support outsourcing of
of Information Science, 31(2), 99–113.
logistics services. Supply Chain Management: An International Journal, 11(4),
Shih, H. S., Shyur, H. J., & Lee, E. S. (2007). An extension of TOPSIS for group decision
294–308.
making. Mathematical and Computer Modelling, 45(7/8), 801–813.
Canós, L., & Liern, V. (2008). Soft computing-based aggregation methods for human
Shyjith, K., Ilangkumaran, M., & Kumanan, S. (2008). Multi-criteria decision-making
resource management. European Journal of Operational Research, 189(3),
approach to evaluate optimum maintenance strategy in textile industry. Journal
669–681.
of Quality in Maintenance Engineering, 14(4), 375–386.
Carr, N. G. (2003). IT doesn’t matter. Harvard Business Review, 5, 41–49.
5008 A. Kelemenis, D. Askounis / Expert Systems with Applications 37 (2010) 4999–5008

Siler, W., & Buckley, J. J. (2005). Fuzzy expert systems and fuzzy reasoning. New Jersey, Ward, S. (1999). Information professionals for the next millennium. Journal of
NJ: Wiley. Information Science, 25(4), 239–247.
Wade, M., & Hulland, J. (2004). Review: The resource-based view and information Willcoxson, L., & Chatham, R. (2006). Testing the accuracy of the IT stereotype:
systems research: Review, extension, and suggestions for future research. MIS Profiling IT managers’ personality and behavioural characteristics. Information
Quarterly, 28(1), 107–142. and Management, 43(6), 697–705.
Wang, J. W., Cheng, C. H., & Huang, K. C. (2009). Fuzzy hierarchical TOPSIS for Yager, R. R. (1980). On a general class of fuzzy connectives. Fuzzy Sets and Systems,
supplier selection. Applied Soft Computing, 9(1), 377–386. 4(3), 235–242.
Wang, Y. M., & Elhag, T. M. S. (2006). Fuzzy TOPSIS method based on alpha level sets Yao, J. S., & Wu, K. (2000). Ranking fuzzy numbers based on decomposition principle
with an application to bridge risk assessment. Expert Systems with Applications, and signed distance. Fuzzy Sets and Systems, 116(22), 275–288.
31(2), 309–319. Zanakis, S. H., Solomon, A., Wishart, N., & Dublish, S. (1998). Multi-attribute decision
Ward, J., & Peppard, J. (1996). Reconciling the IT/business relationship: A troubled making: A simulation comparison of select methods. European Journal of
marriage in need of guidance. Journal of Strategic Information Systems, 5(1), Operational Research, 107(3), 507–529.
37–65.