Ewa Roszkowska

MULTI-CRITERIA DECISION MAKING MODELS

BY APPLYING THE TOPSIS METHOD TO CRISP
AND INTERVAL DATA

Abstract
In this paper, one of the multi-criteria models in making decision, a Technique
for Order Preference by Similarity to an Ideal Solution (TOPSIS), is described. Some
of the advantages of TOPSIS methods are: simplicity, rationality, comprehensibility,
good computational efficiency and ability to measure the relative performance for each
alternative in a simple mathematical form.
The paper has a review character. It systematises the knowledge within the scope
of techniques of decision taking with the use of the TOPSIS method. Simple numerical
examples that reference real situations show practical applications of different aspects
of this method.
The paper is organized as follows. The Introduction presents a short overview
of the decision making steps as well as MCDM techniques. Section 1 presents matrix
representation of the MCDM problem. Section 2 describes the TOPSIS procedure
for crisp data, and Section 3 for interval data. The TOPSIS algorithm in group decision
environment in the case of crisp and interval data is also presented. In Section 4
the problem of qualitative data in TOPSIS model is discussed. The numerical examples
showing applications of those techniques in the negotiation process are presented
in Section 5. Finally, conclusions and some concluding remarks are made in last
section.

Keywords
TOPSIS method, numerical data, interval data, positive ideal solution, negative
ideal solution.

Introduction
Multi-criteria decision making (MCDM) refers to making choice of the
best alternative from among a finite set of decision alternatives in terms
of multiple, usually conflicting criteria. The main steps in multi-criteria decision
making are the following [Hwang, Yoon, 1981; Jahanshahloo, Hosseinzadeh,
Lofti, Izadikhah, 2006a]:
 MULTI-CRITERIA DECISION MAKING MODELS... 201

– establish system evaluation criteria that relate system capabilities to goals,

– develop alternative systems for attaining the goals (generating alternatives),
– evaluate alternatives in terms of criteria,
– apply one of the normative multiple criteria analysis methods,
– accept one alternative as “optimal” (preferred),
– if the final solution is not accepted, gather new information and go to the
next iteration of multiple criteria optimization.
Multi-criteria decision making techniques are useful tools to help
decision maker(s) to select options in the case of discrete problems. Especially,
with the help of computers, those methods have become easier for the users,
so they have found great acceptance in many areas of decision making
processes in economy or management. Among many multi-criteria techniques,
MAXMIN, MAXMAX, SAW, AHP, TOPSIS, SMART, ELECTRE are the
most frequently used methods [Chen, Hwang, 1992]. The nature of the
recommendations of one of those methods depends on the problem being
addressed: choosing, ranking or sorting. The selection of models/techniques
can be also based on such evaluation criteria as:
– internal consistency and logical soundness,
– transparency,
– ease of use,
– data requirements are consistent with the importance of the issue being
considered,
– realistic time and manpower resource requirements for the analytical
process,
– ability to provide an audit trail,
– software availability, where needed.
The classification methods can be categorized by the type of information
from the decision maker (no information, information on attributes or
information on alternatives), data type or by solution aimed at [Chen, Hwang,
1992, p.16-25]. The MAXMIN technique assumed that the overall performance
of an alternative is determined by its weakest attribute, in the MAXMAX
technique an alternative is selected by its best attribute value. The SAW (Simple
Additive Weighting) method multiplies the normalized value of the criteria
for the alternatives with the importance of the criteria and the alternative
with the highest score is selected as the preferred one. The TOPSIS (Technique
for Order Preference by Similarity to the Ideal Solution) selects the alternative
closest to the ideal solution and farthest from the negative ideal alternative.
The classical TOPSIS method is based on information on attribute from
decision maker, numerical data; the solution is aimed at evaluating, prioritizing
and selecting and the only subjective inputs are weights. The AHP (The Analy-
202 Ewa Roszkowska

tical Hierarchy Process) uses a hierarchical structure and pairwise comparisons.

An AHP hierarchy has at least three levels: the main objective
of the problem at the top, multiple criteria that define alternatives in the middle
and competing alternatives at the bottom. The major weaknesses of TOPSIS
are that it does not provide for weight elicitation, and consistency checking
for judgments; on the other hand, the use of AHP has been significantly
restrained by the human capacity for the information process. From this point
of view, TOPSIS alleviates the requirement of paired comparisons and the
capacity limitation might not significantly dominate the process. Hence,
it would be suitable for cases with a large number of criteria and alternatives,
and especially where objective or quantitative data are given [Shih, Shyur, Lee,
2007]. SMART (The Simple Multi Attribute Rating Technique) is similar
to AHP, a hierarchical structure is created to assist in defining a problem and in
organizing criteria. However, there are some significant differences between
those techniques: SMART uses a different terminology. For example,
in SMART the lowest level of criteria in the value tree (or objective hierarchy)
are called attributes rather than sub-criteria and the values of the standardized
scores assigned to the attributes derived from value functions are called ratings.
The difference between a value tree in SMART and a hierarchy in AHP is that
the value tree has a true tree structure, allowing one attribute or sub-criterion
to be connected to only one higher level criterion. SMART does not use
a relative method for standardizing raw scores to a normalized scale. Instead,
a value function explicitly defines how each value is transformed to the
common model scale. The value function mathematically transforms ratings
into a consistent internal scale with lower limit 0 and upper limit 1.
The ELECTRE (Elimination and Choice Expressing Reality) method was to
choose the best action(s) from a given set of actions, but it can also be applied
to three main problems: choosing, ranking and sorting. There are two main parts
to an ELECTRE application: first, the construction of one or several outranking
relations, which aims at comparing in a comprehensive way each pair
of actions; second, an exploitation procedure that elaborates on the re-
commendations obtained in the first phase.
This paper is focused on the TOPSIS method, which was presented by
Hwang and Yoon [1981] and developed later by many authors [Jahanshahloo,
Lofti, Izadikhah, 2006a; 2006b; Zavadskas,Turskis, Tamosaitiene, 2008; Hung,
Chen, 2009]. The acronym TOPSIS stands for Technique for Order Preference
by Similarity to the Ideal Solution. It is worth noting that the TOPSIS method
corresponds to the Hellwig taxonomic method of ordering objects [Hellwig,
1968]. The main advantages of this method are the following [Hung, Cheng,
2009]:
– simple, rational, comprehensible concept,
– intuitive and clear logic that represent the rationale of human choice,
– ease of computation and good computational efficiency,
 MULTI-CRITERIA DECISION MAKING MODELS... 203

– a scalar value that accounts for both the best and worst alternatives ability
to measure the relative performance for each alternative in a simple
mathematical form,
– possibility for visualization.
In general, the process for the TOPSIS algorithm starts with forming
the decision matrix representing the satisfaction value of each criterion with
each alternative. Next, the matrix is normalized with a desired normalizing
scheme, and the values are multiplied by the criteria weights. Subsequently,
the positive-ideal and negative-ideal solutions are calculated, and the distance
of each alternative to these solutions is calculated with a distance measure.
Finally, the alternatives are ranked based on their relative closeness to the ideal
solution. The TOPSIS technique is helpful for decision makers to structure
the problems to be solved, conduct analyses, comparisons and ranking of the
alternatives. The classical TOPSIS method solves problems in which all
decision data are known and represented by crisp numbers. Most real-world
problems, however, have a more complicated structure. Based on the original
TOPSIS method, many other extensions have been proposed, providing support
for interval or fuzzy criteria, interval or fuzzy weights to modeled imprecision,
uncertainty, lack of information or vagueness.
In this paper, the classical TOPSIS algorithms for crisp, as well as
interval data are described. Interval analysis is a simple and intuitive way
to introduce data, uncertainty for complex decision problems, and can be used
for many practical applications. An extension of the TOPSIS technique to
a group decision environment is also investigated. The context of multi-criteria
group decision making in both crisp and interval data are described. Finally,
situations where criteria and their weight are subjectively expressed by
linguistic variables are considered. The practical applications of the TOPSIS
technique in estimating offers, for instance, in buyer-seller exchange are also
proposed.

1. The matrix representation of the MCDM problem

The MCDM problems can be divided into two kinds. One is the classical
MCDM set of problems among which the ratings and the weights of criteria are
measured in crisp numbers. Another one is the multiple criteria decision-making
set of problems where the ratings and the weights of criteria evaluated
on incomplete information, imprecision, subjective judgment and vagueness
are usually expressed by interval numbers, linguistic terms, fuzzy numbers
or intuitive fuzzy numbers.
In the classical MCDM model, we assume exact data, objective and
precise information, but this is often inadequate to model real life situations.
Human judgments are often vague under many conditions. The socio-economic
204 Ewa Roszkowska

environment becomes more complex, the preference information provided by

decision-makers is usually imprecise, and can create hesitation or uncertainty
about preferences. A decision may have to be made under time pressure
and lack of knowledge or data, or the decision-makers have limited attention
and information processing capacities. Most input information is not known
precisely, so that the values of many criteria are expressed in subjective
or uncertain terms. The criteria, as well as their weight, could be subjectively
expressed by linguistic variables. Thus, many researchers extended the MCDM
approach for decision making problems with subjective criteria, interval data
or fuzzy environment using grey system theory or fuzzy set theory.
The grey system theory, developed by Deng [1982, 1988] is based upon
the concept that information is sometimes incomplete or unknown [Jadidi,
Hong, Firouzi, Yusuff, 2008; Liu, Lin, 2006]. Exactly, the theory is based
on the degree of information known which is modeled by intervals. If the
system information is unknown, it is called a black system, if the information
is fully known, it is called a white system. And a system with information
known partially is called a grey system. The fuzzy set theory cannot handle
incomplete data and information, but is adequate to deal with uncertain
and imprecise data [Kahraman, 2008; Chen, Hwang, 1992]. The advantage
of the grey theory over the fuzzy theory is that the grey theory takes into
account the condition of the fuzziness; that is, the grey theory can deal flexibly
with the fuzziness situation.
We can also consider single decision making and group decision making.
Group decision making is more complex than single decision making because
it involves many contradicting factors, such as: conflicting individual goals,
inefficient knowledge, validity of information, individual motivation, personal
opinion, power. In both multi-criteria decision making (MCDM) and group
decision making (GDM), there are two steps: aggregation and exploitation.
In MCDM, aggregation consists in combining satisfaction over different criteria
while GDM problem consists in combining the experts’ opinions into a group
collective one. Group decision making can be approached from two points
of view. In the first approach, individual multi-criteria models are developed
based on individuals’ preferences. Each decision maker formulates a multi-
-criteria problem defining the parameters according to these preferences and
solves the problem getting an individual solution set. Next, the separate
solutions are aggregated by aggregation of operations resulting in the group
solution. In the second approach, each decision maker provides a set
of parameters which are aggregated by appropriate operators, providing finally
a set of group parameters. Upon this set the multi-criteria method is applied
and the solution expresses group preference [Rigopoulos, Psarras, Askounis,
2008].
 MULTI-CRITERIA DECISION MAKING MODELS... 205

Solving of each multi-criteria problem (individual or group decision)

begins with the construction of a decision making matrix (or matrices). In such
matrixes, values of the criteria for alternatives may be real, intervals numbers,
fuzzy numbers or qualitative labels.
Let us denote by D = {1, 2, ... , K } a set of decision makers or experts.
The multi-criteria problem can be expressed in k − matrix format in the
following way:

C1 C2 … Cn
A1 x11k x12k … x1kn
A2 k
x 21 k
x 22 … x 2kn
… … … … …
Am x k
x k … k
x mn
m1 m2

where:
− A1 , A2 , ... , Am are possible alternatives that decision makers have to choose
from,
− C1 ,C2 , ... , Cn are the criteria for which the alternative performance is
measured,
− xijk is the k − decision maker rating of alternative A with respect to the
i
k
criterion C j ( xij is numerical, interval data or fuzzy number).
In this way for m alternatives and n criteria we have matrix X k = xijk ( )
where xijk is value of i − alternative with respect to j − criterion for k −
decision maker, j = 1,2, ... , n, k = 1,2, ... , K .
The relative importance of each criterion is given by a set of weights
which are normalized to sum to one. Let us denote by W k = [ w1k , w2k , ... , wnk ]
a weight vector for k − decision maker, where wkj ∈ ℜ is the k − decision
maker weight of criterion C j and w1k + w2k + ... + wnk = 1.
In the case of one decision maker we write xij , w j , X, respectively.
Multi-criteria analysis focuses mainly on three types of decision
problems: choice − select the most appropriate (best) alternative, ranking −
draw a complete order of the alternatives from the best to the worst, and
sorting − select the best k alternatives from the list.
206 Ewa Roszkowska

2. The classical TOPSIS method

In the classical TOPSIS method we assume that the ratings of alternatives
and weights are represented by numerical data and the problem is solved
by a single decision maker. Complexity arises when there are more than one
decision makers because the preferred solution must be agreed on by interest
groups who usually have different goals. The classical TOPSIS algorithm
for a single decision maker and for group decision making is systematically
described in Section 2.1 and Section 2.2, respectively.

2.1. The classical TOPSIS method for a single decision maker

The idea of classical TOPSIS procedure can be expressed in a series

of following steps [Chen, Hwang, 1992; Jahanshahloo, Lofti, Izadikhah, 2006a].

Step 1. Construct the decision matrix and determine the weight of criteria.
( )
Let X = xij be a decision matrix and W = [ w1 , w2 , ... , wn ] a weight vector,
where xij ∈ ℜ , w j ∈ ℜ and w1 + w2 + ... + wn = 1.
Criteria of the functions can be: benefit functions (more is better) or cost
functions (less is better).

Step 2. Calculate the normalized decision matrix.

This step transforms various attribute dimensions into non-dimensional
attributes which allows comparisons across criteria. Because various criteria
are usually measured in various units, the scores in the evaluation matrix X
have to be transformed to a normalized scale. The normalization of values
can be carried out by one of the several known standardized formulas. Some
of the most frequently used methods of calculating the normalized value nij
are the following:
xij
nij = ,
m (2.1)
∑ i =1
xij2

xij
nij = , (2.1*)
max xij
i
 MULTI-CRITERIA DECISION MAKING MODELS... 207

⎧ xij − min xij

⎪ i
if Ci is a benefit criterion
⎪⎪ max xij − min xij
nij = ⎨
i i (2.1**)
⎪ max xij − xij
i
⎪ max x − min x if Ci is a cos t criterion
⎪⎩ i ij i
ij

for i = 1, … , m; j = 1, … , n.

Step 3. Calculate the weighted normalized decision matrix.

The weighted normalized value vij is calculated in the following way:

vij = w j nij for i = 1, … , m; j = 1, … , n. (2.2)

n
where w j is the weight of the j-th criterion, ∑w
j =1
j = 1.

Step 4. Determine the positive ideal and negative ideal solutions.

Identify the positive ideal alternative (extreme performance on each criterion)
and identify the negative ideal alternative (reverse extreme performance on each
criterion). The ideal positive solution is the solution that maximizes the benefit
criteria and minimizes the cost criteria whereas the negative ideal solution
maximizes the cost criteria and minimizes the benefit criteria.
Positive ideal solution A + has the form:
⎛ ⎛ ⎞⎞
⎜⎛ ⎞⎜ ⎟⎟
+
A = (
v1+ , v2+ ,..., vn+ ) = ⎜ ⎜ maxvij | j ∈ I ⎟, ⎜ minvij | j ∈ J ⎟ ⎟. (2.3)
⎜ ⎜⎝ i ⎟ i
⎠⎜ ⎟⎟
⎝ ⎝ ⎠⎠

Negative ideal solution A − has the form:

⎛ ⎛ ⎞⎞
⎜⎛ ⎞⎜ ⎟⎟
−
A = (
v1− , v2− ,..., vn− ) ⎜ ⎟
= ⎜ minvij | j ∈ I , ⎜ maxvij | j ∈ J ⎟ ⎟ (2.4)
⎜ ⎜⎝ i ⎟ i
⎠⎜ ⎟⎟
⎝ ⎝ ⎠⎠
where I is associated with benefit criteria and J with the cost criteria,
i = 1, … , m; j = 1, … , n.
208 Ewa Roszkowska

Step 5. Calculate the separation measures from the positive ideal solution
and the negative ideal solution.
In the TOPSIS method a number of distance metrics can be applied*.
The separation of each alternative from the positive ideal solution is given as
1/ p
⎛ ⎞
∑( )⎟
n
+ p⎟
d i+ =⎜ vij − v j , i = 1,2, … , m. (2.5)
⎜
⎝ j =1 ⎠
The separation of each alternative from the negative ideal solution is given as
1/ p
⎛ n ⎞
∑( ) p
d i− =⎜ v − v −j ⎟ , i = 1,2, … , m. (2.6)
⎜ j=1 ij ⎟
⎝ ⎠
Where p ≥ 1 . For p = 2 we have the most used traditional n-dimensional
Euclidean metric.

∑ (v )
n
2
d i+ = ij − v+j , i = 1,2, … , m, (2.5*)
j =1

∑ (v )
n
2
d i− = − v −j , i = 1,2, … , m. (2.6*)
ij
j =1

Step 6. Calculate the relative closeness to the positive ideal solution.

The relative closeness of the i-th alternative Aj with respect to A+ is defined as
d i−
Ri = , (2.7)
d i− + d i+

where 0 ≤ Ri ≤ 1, i = 1,2, … , m.

Step 7. Rank the preference order or select the alternative closest to 1.

A set of alternatives now can be ranked by the descending order of the value
of Ri .

*
Possible metrics the first power metric (the least absolute value terms), Tchebychev metric or others [see
Kahraman, Buyukozkan, Ates, 2007; Olson 2004].
 MULTI-CRITERIA DECISION MAKING MODELS... 209

2.2. The classical TOPSIS method for group decision making

In this part we explain the detailed TOPSIS procedure for group decision
making based on the Shih, Shyur and Lee proposition [Shih, Shyur, Lee, 2007].

Step 1. Construct the decision matrixes and determine the weights

of criteria for k-decision makers.
Let X k = xijk ( ) be a decision matrix, W k = [ w1k , w2k , ... , wnk ] weight vector
for k − decision maker or expert, where xijk ∈ ℜ , wkj ∈ ℜ , w1k + w2k + ... + wnk = 1
for k = 1,2, ... , K .

Step 2. Calculate the normalized decision matrix for each decision maker.
In this step some of the earlier described methods of normalization can be used.
Let us assume that we use
xijk
rijk = .
(2.8)
∑( )
m
2
xijk
i =1

In this procedure weights are manipulated in the next step.

Step 3. Determine the positive ideal and negative ideal solutions for each
decision maker.
The positive ideal solution A k + for k − decision maker has the form
⎧ ⎛ ⎞⎫
⎪⎛⎜ ⎞⎜ ⎟⎪
A k+
= {
r1k + , r2k + ,..., rnk + } = ⎨ max(rij ) | j ∈ I ⎟, ⎜ min(rij ) | j ∈ J ⎟ ⎬.
⎜
k
⎟
k (2.9)
⎪⎝ i ⎠⎜ i ⎟⎪
⎩ ⎝ ⎠⎭

The negative ideal solution A k − for k- decision maker has the form:
⎧ ⎛ ⎞⎫
⎪⎛⎜ ⎞⎜ ⎟⎪
A k−
= {
r1k − , r2k − ,..., rnk − } = ⎨ min (rij ) | j ∈ I ⎟, ⎜ max(rij ) | j ∈ J ⎟ ⎬,
⎜
k
⎟
k (2.10)
⎪⎝ i ⎠⎜ i ⎟⎪
⎩ ⎝ ⎠⎭
where I is associated with the benefit criteria and J with the cost criteria.
210 Ewa Roszkowska

Step 4. Calculate the separation measures from the positive ideal solution
and the negative ideal solution.

Step 5.1. Calculate the separatation measure for individuals.

The separation of i-th alternative Ai from the positive ideal solution Ak +
for each k − decision maker is given as
1
⎛ ⎞ p
∑ ( )
m
p
d ik+ =⎜ wkj rijk − r jk+ ⎟ , i = 1,2, … , m. (2.11)
⎜ ⎟
⎝ j=1 ⎠

The separation of i-th alternative Ai from the negative ideal solution Ak −

for each k − decision maker is given as
1
⎛ ⎞ p
∑ ( )
m
d ik − =⎜ wkj rijk − r jk −
p
⎟ , i = 1,2, … , m, (2.12)
⎜ ⎟
⎝ j=1 ⎠
where p ≥ 1 . For p = 2 we have the Euclidean metric.

Step 5.2. Calculate the separation measure for the group.

The aggregation for measure for the group measures of the positive ideal di*+
and negative ideal solution d i*− for the i-th alternative Ai is given by one of the
operators:
arithmetic mean:
K K

∑ dik+ ∑d k−
i
(2.13)
di*+ = k=1
and di*− = k=1
K K

or
geometric mean:
K K
di*+ = K ∏dik+ and di*− = K ∏d k−
i . (2.13*)
k=1 k=1

Step 6. Calculate the relative closeness to the positive ideal solution.

The relative closeness of the alternative Ai to the positive ideal solution
is defined as
 MULTI-CRITERIA DECISION MAKING MODELS... 211

d i*−
Ri* = for i = 1,2, … , m (2.14)
d i*− + d i*+
where 0 ≤ Ri* ≤ 1.
The larger the index value, the better the evaluation of the alternative.
Step 7. Rank the preference order or select the alternative closest to 1.
A set of alternatives can now be ranked by the descending order of the value
of Ri* .

3. The TOPSIS method with criteria values

determined as interval
In some cases determining the exact value of criteria is difficult
and decision makers are usually more comfortable providing intervals to specify
model input parameters. An interval number data formulation is a simple
and intuitive way to represent uncertainty, which is typical of real decision
problems. Here, the TOPSIS method using interval as the basis for evaluating
value alternatives is described. However, we can also consider an interval
weights description [Jadidi, Hong, Firouzi, Yusuff, 2008].

3.1. The TOPSIS method with attributed values determined

as interval for a single decision maker
An algorithmic method which extends TOPSIS for decision-making
problems with interval data was proposed by Jahanshahloo, Lofti, Izadikhah.
This procedure can be described in the following steps [Jahanshahloo, Lofti,
Izadikhah, 2006a].
Step 1. Construct the decision matrix and determine the weight of criteria.
( )
Let X = xij be a decision matrix and W = [ w1 , w2 , ... , wn ] a weight vector,
[
where xij = x ij , x ij ] , xij , x ij ∈ ℜ , w j ∈ ℜ and w1 + w2 + ... + wn = 1.

Step 2. Calculate the normalized interval decision matrix.

The normalized values n ij , n ij are calculated in the following way:
x ij
n ij = for i = 1, … , m; j = 1, … , n.
∑ ((x ) + (x ) )
m
(3.1)
2 2
ij ij
i=1
212 Ewa Roszkowska

x ij
n ij = for i = 1, … , m; j = 1, … , n. (3.2)
m

i=1
(
∑ (x ij ) + (x ij )
2 2
)
[ ]
The interval nij , nij is normalized value of interval x ij , x ij . [ ]
Step 3. Calculate the weighted normalized interval decision matrix.
The weighted normalized values v ij and v ij are calculated in the
following way:
v ij = w j n ij for i = 1, …, m; j = 1, … , n, (3.3)

v ij = w j n ij for i = 1, …, m; j = 1, … , n, (3.4)
n
where w j is the weight of the j-th criterion, ∑w j =1
j = 1.

Step 4. Determine the positive ideal and negative ideal solutions.

The positive ideal solution has the form A+ :
⎛ ⎛ ⎞⎞
⎜⎛ ⎞⎜ ⎟⎟
+
A = ( v1+ , v2+ , ... , vn+ ) = ⎜ ⎜⎜ maxv ij | j ∈ I ⎟⎟, ⎜ minv ij | j ∈ J ⎟ ⎟ . (3.5)
⎜⎜ ⎝ i ⎠⎜ i ⎟ ⎟⎟
⎝ ⎝ ⎠⎠
The negative ideal solution has the form A − :
⎛ ⎞⎞
⎜⎛ ⎞ ⎛⎜ ⎟⎟
( ⎜
)
A− = v1− , v2− , ... , vn− = ⎜ ⎜ min v ij | j ∈ I ⎟, ⎜ max v ij | j ∈ J ⎟ ⎟ ,
⎟⎜ i
(3.6)
⎜⎝ i ⎠⎝ ⎟⎟
⎝ ⎠⎠
where I is associated with benefit criteria and J with cost criteria.

Step 5. Calculate the separation measures from the positive ideal solution
and the negative ideal solution.
The separation of each alternative from the positive ideal solution is given as*:

∑ (v ) ∑ (v )
n n
2 2
d i+ = +
− v+j (3.7)
ij − v j + ij , i = 1,2, … , m.
j =1 j =1

*
Traditional TOPSIS applied to Euclidean norm is presented here. However, we can also use other metrics.
 MULTI-CRITERIA DECISION MAKING MODELS... 213

The separation of each alternative from the negative ideal solution is given as:

∑ (v ) ∑ (v )
n n
2 2
d i− = −
ij − v j + ij − v −j , i = 1,2, … , m. (3.8)
j =1 j =1

Step 6. Calculate the relative closeness to the positive ideal solution.

The relative closeness of the alternative Ai with respect to A+ is defined as:
d i− (3.9)
Ri = for i = 1,2, … , m,
d i− + d i+
where 0 ≤ Ri ≤ 1.

Step 7. Rank the preference order or select the alternative closest to 1.

The set of alternatives can now be ranked by the descending order of the value
of Ri .

3.2. The TOPSIS method with attributed values determined

as intervals for group decision making

We assume here that values are considered as intervals and we have

group of k − decision makers [Zavadskas, Turskis, Tamosaitiene, 2008].

Step 1. Construct the decision matrixes and determine the weights

of criteria for k − decision makers.
( )
Let X k = xijk be a decision matrix, W k = [ w1k , w2k , ... , wnk ] is weight vector

xijk = ⎡ x ij , xij ⎤ ,
k k k k
for k − decision maker, where x ij , xij ∈ ℜ , wi ∈ ℜ ,
⎢⎣ ⎥⎦
w1k + w2k + ... + wnk = 1 for k = 1,2, ... , K .

Step 2. Calculate the normalized grey decision matrixes.

This step transforms various attribute dimensions into non-dimensional
attributes rijk = ⎡r ij , r ij ⎤ , which allows comparisons across criteria.
k k
⎢⎣ ⎥⎦
k k
The normalized values r ij , r ij are calculated in the following way:
214 Ewa Roszkowska

k k
k x ij k x ij
r =
ij k
, r =
ij k (3.10)
max ( x ij ) max ( x ij )
i i

for i = 1, … , m; j = 1, … , n, k = 1,2, … , K.

Step 3. Determine the positive ideal and the negative ideal solutions
for each decision maker.
The positive ideal solution Ak + for k-decision maker has the following
form:
⎧ ⎛ ⎞⎫
⎞⎜ ⎟⎪
A k+
= {
r1k + , r2k + , ... , rnk + } ⎪⎛ k k
= ⎨⎜⎜ max(r ij ) | j ∈ I ⎟⎟, ⎜ min(r ij ) | j ∈ J ⎟⎬. (3.11)
⎪⎝ i ⎠⎜ i ⎟⎪
⎩ ⎝ ⎠⎭
The negative ideal solution Ak − for k-decision maker has the form:
⎧ ⎞⎫
⎞ ⎛⎜
A k−
= {r k− k−
1 , r2 , ... , rn
k−
}= ⎪⎨⎛⎜⎜ min (r k
ij ) |
⎟ k ⎟⎪
j ∈ I , ⎜ max ( r ij ) | j ∈ J ⎟ ⎬,
⎟⎜ i
(3.12)
⎪⎝ i ⎠⎝ ⎟⎪
⎩ ⎠⎭
where I is associated with benefit criteria and J with cost criteria.

Step 4. Calculate the separation measures from the positive ideal solution
and the negative ideal solution.

Step 5.1. Calculate the separation measure for individuals.

The separation of i-th alternative Ai from the positive ideal solution Ak +
for each k − decision maker is given as
1/ p
⎛ ⎞ ⎞⎟
⎛
( )
m p
1
∑ k p
+ ⎛⎜ r jk + − r ij ⎞⎟
k
dik + = ⎜ wkj ⎜⎜ rjk + − r ij ⎟⎟ , i = 1,2, … , m. (3.13)
⎜2 ⎝ ⎝ ⎠ ⎠ ⎟⎠
⎝ j=1

The separation of i-th alternative Ai from the negative ideal solution Ak −

for each k − decision maker is given as

( )
1/ p

d i
k− ⎛1 m
(
= ⎜ ∑ w kj ⎛⎜ rjk − − r ij
k
) p
+r k−
j −r
k p
ij
⎞⎞
⎟⎟ , i = 1,2, … , m. (3.14)
⎝ 2 j=1 ⎝ ⎠⎠
 MULTI-CRITERIA DECISION MAKING MODELS... 215

If p = 2, then the metric is a weighted grey number Euclidean distance function.

Equations (3.13) and (3.14) will be as follows:

d ik + =
1 m k ⎛ k+
2 j=1 ⎝
( k 2
) (
k 2⎞
)
∑ w j ⎜ rj − r ij + rjk + − r ij ⎟ , i = 1,2, … , m.
⎠
(3.13*)

The separation of i-th alternative Ai from the negative ideal solution Ak −

is given as

d ik − =
1 m k ⎛ k−
2 j=1 ⎝
( k 2
) (
k 2⎞
)
∑ w j ⎜ rj − r ij + rjk − − r ij ⎟ , i = 1,2, … , m.
⎠
(3.14*)

Step 5.2. Calculate the separation measure for the group.

The aggregation of the measure for the group measures of the positive ideal di*+
and the negative ideal solution d i*− for the i-th alternative Ai is given by:
arithmetic mean:
K K

∑ dik+ ∑d k−
i
(3.15)
di*+ = k=1 and di*− = k=1 ,
K K
or
geometric mean:
K K
di*+ = K ∏
k=1
dik+ and di*− = K ∏d
k=1
k−
i . (3.15*)

Step 6. Calculate the relative closeness to the positive ideal solution.

The relative closeness of the alternative Ai with respect to A+ is defined as
d *−
Ri* = *− i *+ for i = 1,2, … , m, (3.16)
d i + di
where 0 ≤ Ri* ≤ 1.
The larger the index value, the better the evaluation of alternative.

Step 7. Rank the preference order or select the alternative closest to 1.

The set of alternatives can now be ranked by the descending order of the value
of Ri* .
216 Ewa Roszkowska

4. The quantitative and qualitative criteria

in the TOPSIS method. Weights expressed
by linguistic variable
In the TOPSIS algorithm the quantitative criteria are scaled using their
own real numbers and for representation of the imprecision of spatial data,
and human cognition over the criteria of the theory of linguistic variables
is used. A linguistic variable is a variable where values are words or sentences
in a natural or artificial language. Especially, since traditional quantification
methods are difficult to describe situations that are overtly complex or hard
to describe, the notion of a linguistic variable is necessary and useful. We can
use this kind of expression for rating qualitative criteria as well as to compare
two evaluation criteria.
The qualitative criterion can be described using linguistic variables; next
the criteria ratings on the 1-9 number scale (Table 1) or on the 1-7 interval scale
(Table 2) can be provided, respectively [Jadidi, Hong, Firouzi, Yusuff, Zulkifli,
2008].
Table 1
The scale of alternative ratings for qualitative criterion in the case of classical TOPSIS method
Scale Rating
Poor (P) 1
Medium poor (MP) 3
Fair (F) 5
Medium good (MG) 7
Good (G) 9
Intermediate values between the two adjacent 2,4,6,8
judgments

Table 2

The scale of alternative ratings in the case of interval TOPSIS method

Scale Rating
Very Poor (VP) [0,1]
Poor (P) [1,3]
Medium poor (MP) [3,4]
Fair (F) [4,5]
Medium good (MG) [5,6]
Good (G) [6,9]
Very good (VG) [9,10]
 MULTI-CRITERIA DECISION MAKING MODELS... 217

Each decision maker individually uses linguistic variables transformed

for numerical scale (Table 1) or interval scale (Table 2) to identify
the alternative rankings for the subjective criterion. Then the rating value
for group decision makers can be calculated using the following formula

xij =
1 1
K
[ ]
xij + xij2 + ... + xijK , (4.1)

where:
xijs − is the rating value of alternative Ai with the respect to quantitative
criterion C j (crisp or interval) of s − decision maker (i = 1,2, ... , n;
j = 1,2, ... , m; s = 1,2, ... , K ).

In this way for m alternatives and n criteria and K − decision makers

we can obtain one aggregated matrix X = (xij ) where x ij is value
of i − alternative with respect to j − criterion for i = 1,2, ... , m; j = 1, 2, ... , n.
The weights of the factors are subjective input, which directly influences
the final evaluated result. In the evaluating system, the importance of every
index is, in general, not equal, so they must be set different weight factors.
Among many ways to set the weight factors are for instance the Delphi method
or the AHP method [Olson, 2004]. The Delphi method is the most popular
expert evaluating technique. The Delphi method is a forecasting and evaluating
method both qualitative and quantitative which collects experts’ ideas
anonymously, exchanges and corrects this information many times to reach
a consistent idea and gives the subject a final evaluation according to
the experts’ ideal. AHP (The Analytical Hierarchy Process) uses a hierarchical
structure and pairwise comparisons. An AHP hierarchy has at least three levels:
the main objective of the problem at the top, multiple criteria that define
alternatives in the middle and competing alternatives at the bottom. The AHP
method uses system analysis and continuously decomposes the evaluating
indices according to the main evaluating indices of every level [Saaty, 1980].
The classical TOPSIS method does not consider a hierarchical structure
consisting of main attributes and subattributes. This method evaluates
the alternatives with respect to main attributes only with a single level.
The common property of these methods is their ease of implementation, so this
method is often used to obtain weight criteria.
In the case where the criterion weights are linguistic variables,
the weights can be expressed by the 1-9 scale shown in Table 3 [Jadidi, Hong,
Firouzi, Yusuff, Zulkifli, 2008].
218 Ewa Roszkowska

Table 3

The scale of criterion weights

Scale Weight
Very very low (VVL) 0,005
Very low (VL) 0,125
Low (L) 0,175
Medium low (ML) 0,225
Medium (M) 0,275
Medium hight (MH) 0,325
Hight (H) 0,375
Very Hight (VH) 0,425
Very Very Hight (VVH) 0,475

The vector of attribute weights must sum up to 1; if not, it is normalized.

Each decision maker individually uses linguistic variables (Table 3) to identify
the criterion weights. Then the criterion weights for all decision makers can be
calculated using the following formula

wr =
K
[
1 1
]
wr + wr2 + ... + wrK , (4.2)

where:
wrs − is the weight of r − criterion for s − decision makers ( r = 1,2, ... , n;
s = 1, 2, ... , K ).

5. Practical application
In this section, to demonstrate the calculation process of the approaches
described, two examples are provided.

Example 1.
A firm intends to choose the best offer (or ranking of the offers) from the set
of proposals submitted by potential contractors. Two experts evaluate five
proposals using several criteria. In order to simplify the calculation, only four
criteria are considered: deadline of payment after receipt the goods (in days),
unitary price (in euro), conditions of warranty and contractor reputation,
C1, C2, C3, C4, respectively. The criteria C1, C3, C4 are benefit criteria,
the greater values being better, and C2 is the cost criterion, the smaller values
are better. Criteria C3, C4 are subjectively evaluated by experts on basic
 MULTI-CRITERIA DECISION MAKING MODELS... 219

available information and they are considered as linguistic variables, while

the other criteria are scaled using their own real numbers, respectively. This
is shown in Table 4. Based on Table 1, the decision matrixes for two decision
makers are obtained (Table 5).
Table 4
Criteria rating values for two decision makers

C1 C2 C3 C4
D1
A1 7 21 F&MP MG
A2 7 24 MG&F F&MP
A3 14 25 MP G& MG
A4 14 26 G MP
A5 21 35 MP &F F& MP

D2
A1 7 21 F&MP MG&F
A2 7 24 MG F
A3 14 25 MP&P G &MG
A4 14 26 MG MP&P
A5 21 35 MP MP

Table 5
Decision matrixes for two decision makers

C1 C2 C3 C4
D1
A1 7 21 4 7
A2 7 24 6 4
A3 14 25 3 8
A4 14 26 9 3
A5 21 35 4 4

D2
A1 7 21 4 6
A2 7 24 7 5
A3 14 25 2 8
A4 14 26 7 3
A5 21 35 3 3
220 Ewa Roszkowska

Based on linguistic variables, the evaluation values of attribute weights for

the first and second decision makers can be obtained and the results are shown
in Table 6. The normalized criteria weights for each decision maker obtained
using Table 3 are shown in Table 7.
Table 6

Criteria weights for two decision makers

C1 C2 C3 C4
D1
L VH L ML

D2
ML VVH VL L

Table 7

Normalized criteria weights for two decision makers

C1 C2 C3 C4
D1
0,175 0,425 0,175 0,225

D2
0,225 0,475 0,125 0,175

CASE 1. Rank the preference order for individual decision makers

Using formulas 2.1-2.7 the calculation results on data from Table 5 and Table 7
and rank order for each decision maker are shown in Table 8 and Table 9,
respectively.
Table 8
Calculation results for 1-decision maker
The separation The separation The relative closeness
of alternatives of alternatives of alternatives
Alternatives Rank
to positive ideal to negative ideal to the positive ideal
one one one
A1 0.107805 0.124281 0.535497 2
A2 0.117942 0.090785 0.434945 4
A3 0.096981 0.122181 0.557491 1
A4 0.105379 0.112779 0.516961 3
A5 0.141766 0.083486 0.370633 5
 MULTI-CRITERIA DECISION MAKING MODELS... 221

Table 9

Calculation results for 2-decision maker

The separation The separation The relative closeness
of alternatives of alternatives of alternatives
Alternatives Rank
to positive ideal to negative ideal to the positive ideal
one one one
A1 0.112485 0.128548 0.533321 2
A2 0.115014 0.113042 0.495676 3
A3 0.082214 0.130468 0.613440 1
A4 0.110660 0.104396 0.485436 4
A5 0.141415 0.104894 0.425865 5

CASE 2. Rank the preference order for group decision makers

(1 method)
The decision matrix is calculated using formula (4.1) and attributes weights
of the criteria using (4.2). The results are shown in Table 10 and Table 11,
respectively.

Table 10

Decision matrix for group decision makers

C1 C2 C3 C4
A1 7 21 4 6,5
A2 7 24 6,5 4,5
A3 14 25 2,5 8
A4 14 26 8 2,5
A5 21 35 3,5 3,5

Table 11

Normalized criteria for group decision makers

C1 C2 C3 C4

0.20 0.45 0.15 0.20

Using formulas 2.1-2.7 the calculation results on data based on Table 10

and Table 11 and rank order for group decision making are shown in Table 12.
222 Ewa Roszkowska

Table 12

Calculation results for group decision making (1 method)

The separation The separation The relative closeness
of alternatives of alternatives of alternatives
Alternatives Rank
to positive ideal to negative ideal to the positive ideal
one one one
A1 0.107701 0.126396 0.539930 2
A2 0.112581 0.102844 0.477401 4
A3 0.088631 0.127144 0.589244 1
A4 0.108991 0.107566 0.496711 3
A5 0.141512 0.094110 0.399412 5

CASE 3. Rank the preference order for group decision makers

(2 method)
Using formulas 2.8-2.14 in the case of the Euclidean metric (p = 2)
and arithmetic mean (formula 2.13) the calculation results on data based
on Table 5 and Table 7 and rank order for group decision making are shown
in Table 13.

Table 13

Calculation results
The separation The separation The relative closeness
of alternatives of alternatives of alternatives
Alternatives Rank
to positive ideal to negative ideal to the positive ideal
one one one
A1 0.250522 0.222860 0.470783 3
A2 0.253325 0.199743 0.440868 4
A3 0.210667 0.255295 0.547889 1
A4 0.235766 0.229436 0.493196 2
A5 0.273169 0.211055 0.435863 5

Remark 1. Let us observe that we obtain different rank order in Case 2

and Case 3.

Example 2.
A firm intends to choose the best offer (or ranking of the offers) from the set
of three proposals submitted by potential contractors. As in example 1, two
experts evaluate each proposal using the same four criteria: deadline of payment
 MULTI-CRITERIA DECISION MAKING MODELS... 223

after receipt of the goods (in days), unitary price (in euro), conditions
of warranty and contractor's reputation, C1, C2, C3, C4 respectively. The criteria
C1, C3, C4 are benefit criteria, greater values being better, and C2 is the cost
criterion, smaller values being better. Criteria C3, C4 are subjectively evaluated
by the experts on basic available information and they are considered now
as linguistic variables, and the other criteria are scaled using interval data,
respectively. This is shown in Table 14.

Table 14

The interval decision matrix for two decision makers

C1 C2 C3 C4
1 1
x i1 xi11 xi2 xi21
D1
A1 0 7 20 22 P P
A2 7 14 22 24 G MG
A3 14 21 24 26 MP F

2 2
x i1 xi12 xi2 xi22
D2
A1 0 7 20 22 P MP
A2 7 14 22 24 MP P
A3 14 21 24 26 MP MP

Based on Table 2, the decision matrixes of two decision makers are obtained
(Table 15).

Table 15

The interval decision matrix for two decision makers

C1 C2 C3 C4
1 1 1 1 1 1 1 1
x i1 x i1 xi2 xi2 x i3 x i3 xi4 xi4
D1
A1 0 7 20 22 1 3 1 3
A2 7 14 22 24 6 9 5 6
A3 14 21 24 26 3 4 4 5
224 Ewa Roszkowska

Table 15 contd.

2 2 2 2 2 2 2 2
x i1 x i1 xi2 xi2 x i3 x i3 xi4 xi4
D2
A1 0 7 20 22 1 3 3 4
A2 7 14 22 24 3 4 1 3
A3 14 21 24 26 3 4 3 4

Based on linguistic variables the evaluation values of attribute weight for each
decision maker can be obtained and the results are shown in Table 16.

Table 16

Criteria weights for two decision makers

C1 C2 C3 C4
D1
ML VVH M L

D2
ML VVH VL L

The normalized criteria weights for two decision makers are shown in Table 17.

Table 17

Normalized criterion weights for two decision makers

C1 C2 C3 C4
D1
0.196 0.413 0.239 0.152

D2
0.225 0.475 0.125 0.175

CASE 1: Rank the preference order for individual decision makers

Using formulas 3.1-3.9 the calculation results on data from Table 15
and Table 17 and rank order for each decision maker are shown in Table 18
and Table 19, respectively.
 MULTI-CRITERIA DECISION MAKING MODELS... 225

Table 18

Calculation results for 1-decision maker

The separation The separation The relative closeness
of alternatives of alternatives of alternatives
Alternatives Rank
to positive ideal to negative ideal to the positive ideal
one one one
A1 0.266630 0.084412 0.240462 3
A2 0.121513 0.230392 0.654700 1
A3 0.169564 0.191156 0.529928 2

Table 19

Calculation results for 2-decision maker

The separation The separation The relative closeness
of alternatives of alternatives of alternatives
Alternatives Rank
to positive ideal to negative ideal to the positive ideal
one one one
A1 0.195026 0.118375 0.377710 3
A2 0.141770 0.141991 0.500388 2
A3 0.084299 0.211994 0.715488 1

CASE 2. Rank the preference order for group decision makers

(1 method)
The decision matrix is calculated using formula (4.1) and attributes weights
of the criteria using (4.2). The results are shown in the Table 20 and
the Table 21, respectively.

Table 20

Decision table for group decision making

C1 C2 C3 C4
1 1 1 1 1 1 1 1
x i1 x i1 xi2 xi2 x i3 x i3 xi4 xi4

A1 0 7 20 22 1 3 2 3,5
A2 7 14 22 24 4,5 6,5 3 4,5
A3 14 21 24 26 3 4 3,5 4,5
226 Ewa Roszkowska

Table 21

Normalized criteria weights for group decision makers

C1 C2 C3 C4

0.2105 0.4440 0.1820 0.1635

Using formulas 3.1-3.9 the calculation results on data based on Table 20

and Table 21 and rank order for group decision making are shown in Table 22.

Table 22

Calculation results for group decision making (1 method)

The separation The separation The relative closeness
of alternatives of alternatives of alternatives
Alternatives Rank
to positive ideal to negative ideal to the positive ideal
one one one
A1 0.217917 0.087577 0.286673 3
A2 0.122574 0.172678 0.584850 2
A3 0.110306 0.194664 0.638305 1

CASE 3. Rank the preference order for group decision making

(2 method)
Using formulas 3.10-3.16 in the case of the Euclidean metric (formula
3.13*-3.14*) and arithmetic mean (3.15) the calculations results on data based
on Table 15 and Table 17 and rank order for group decision making are shown
in Table 23.

Table 23

Calculation results for group decision making (2 method)

The separation The separation The relative closeness
of alternatives of alternatives of alternatives
Alternatives Rank
to positive ideal to negative ideal to the positive ideal
one one one
A1 0.224004 0.125446 0.358981 3
A2 0.125382 0.223736 0.640861 2
A3 0.173915 0.322996 0.650008 1

Remark 2. Let us observe that we obtain the same rank order in Case 2
and Case 3.
 MULTI-CRITERIA DECISION MAKING MODELS... 227

Remark 3. The TOPSIS method presents a universal methodology and

a simplified practical model for ordering and choosing offers in buyer-seller
exchange. This indicator system and evaluation model can be used widely in the
area of bargaining process which is usually complex and uncertain. Negotiators
have to consider qualitative issues such as price, time of payments, as well
as quantitative ones such as reputation, power of negotiation, relationships
between sides and so on. Moreover, human thinking is imprecise, lack
of information, imprecision and evaluations are always restricted by some
objective factors. The concept of the TOPSIS method is clear, the calculation
is simple and convenient and the methodology can be extended and adjusted
to specific environments. According to the TOPSIS analysis results, a ne-
gotiator can choose the most effective alternative that is possible to implement.
The decision maker's evaluation could be based on linguistic variables, crisp
or interval data. The example of the practical application proves that this
method is efficient and feasible.

Concluding remarks
There are a variety of multiple criteria techniques to aid selection
in conditions of multiple-criteria problems. One of them is the TOPSIS method,
where the ranking of alternatives is based on the relative similarity to the ideal
solution, which avoids the situation of having the same similarity index to both
positive ideal and negative ideal solutions.
The TOPSIS method is a practical and useful technique for ranking
and selecting alternatives. In this paper we focused mainly on the concept of the
TOPSIS algorithm for crisp and interval data. An extension of the TOPSIS
technique to a group decision environment was also investigated.
The high flexibility of the TOPSIS concept is able to accommodate
further extensions to make best choices in various situations. Practically,
TOPSIS and its modifications are used to solve many theoretical and real-world
problems. In addition, the preferences of more than one decision makers can be
also aggregated into the TOPSIS procedure. The classical TOPSIS have been
extended according to the requirements of different real-world decision making
problems providing support for interval or fuzzy criteria, interval or fuzzy
weights to modeled imprecision, uncertainty, lack of information or vagueness,
such as TOPSIS with interval data, Fuzzy TOPSIS, Fuzzy AHP and TOPSIS
and group TOPSIS.
In the TOPSIS model based on the theory of fuzzy sets the rating of each
alternative is expressed in triangular or trapezoidal fuzzy numbers, the weight
of each criterion is represented by fuzzy or crisp values, and different
228 Ewa Roszkowska

normalization (for instance Euclidean, linear or others) are used*. The

normalized fuzzy numbers can be calculated by using the concept of α-cuts
[Jahanshaholoo, Lofti, Izadikhah, 2006b]. The TOPSIS model based on the
intuitionistic fuzzy set (IFS) allows also to measure the degree of satisfiability
and the degree of non-satisfiability, respectively, of each alternative evaluated
across a set of criteria [Hung, Chen, 2009; Saghafian, Hejazi, 2005].
The hierarchical TOPSIS method is developed to benefit both from
the superiority of the hierarchical structure of AHP and ease of implementation
of TOPSIS method [Kahraman, Buyukozkan, Ates 2007; Chiang, Cheng, 2009].
In Polish literature, among many applications, the TOPSIS method
(to rank objects) and analytical hierarchy process (to calculate weight
of criteria) was employed to assess the socioeconomic development of rural
Wielkopolska seen as a collection of counties [Łuczak, Wysocki, 2006],
the fuzzy TOPSIS method based on α-level sets was employed to assess the
level of people life in chosen counties in Wielkopolska Province [Łuczak,
Wysocki, 2008], TOPSIS methods for crisp and interval data were used
for ordering offers in buyer-seller transactions [Roszkowska 2009].

References
Chen S.J., Hwang C.L. (1992): Fuzzy Multiple Attribute Decision Making: Methods
and Applications. Springer-Verlag, Berlin.
Chiang K.F., Cheng S.W. (2009): Using Analytic Hierarchy Process Method
and Technique for Order Preference by Similarity to Ideal Solution to Evaluate
Curriculum in Department of Risk Management and Insurance. “Journal
of Social Sciences”, 19(1), pp. 1-8.
Deng J.L. (1982): Control Problems of Grey System. Systems and Control Letters,
Vol. 1, No. 5, pp. 288-294.
Deng J.L. (1988): Introduction to Grey System Theory. “Journal of Grey Theory”,
Vol. 1, Iss. 1, pp. 1-24.
Hellwig Z. (1968): Zastosowania metody taksonomicznej do typologicznego podziału
krajów ze względu na poziom rozwoju i strukturę wykwalifikowanych kadr.
“Przegląd Statystyczny”, z. 4, pp. 307-327.
Hung C.C., Chen L.H. (2009): A Fuzzy TOPSIS Decision Making Model with Entropy
Weight under Intuitionistic Fuzzy Environment. Proceedings of the International
Multi-Conference of Engineers and Computer Scientists IMECS, Hong Kong.
Hwang C.L., Yoon K. (1981): Multiple Attribute Decision Making: Methods and
Applications. Springer-Verlag, Berlin.

*
Comparison of fuzzy TOPSIS methods can be find in [Kahraman, Buyukozkan, Ates, 2007].
 MULTI-CRITERIA DECISION MAKING MODELS... 229

Jahanshahloo G.R., Lofti F.H., Izadikhah M. (2006a): An Algorithmic Method to Extend

TOPSIS for Decision Making Problems with Interval Data. “Applied Ma-
thematics and Computation”, 175, pp. 1375-1384.
Jahanshahloo G.R., Lofti F.H., Izadikhah M. (2006b): Extension of the TOPSIS Method
for Decision-Making Problems with Fuzzy Data. “Applied Mathematics and
Computation”, 181, pp. 1544-1551.
Jadidi O., Hong T.S., Firouzi F., Yusuff R.M., Zulkifli N. (2008): TOPSIS and Fuzzy
Multi-Objective Model Integration for Supplier Selection Problem. “Journal
of Achievements in Materials and Manufacturing Engineering of Achievements
in Materials and Manufacturing Engineering”, 31(2), pp. 762-769.
Jadidi O., Hong T.S., Firouzi F., Yusuff R.M. (2008): An Optimal Grey Based
Approach Based on TOPSIS Concept for Supplier Selection Problem.
“International Journal of Management Science and Engineering Management”,
Vol. 4, No. 2, pp. 104-117.
Kahraman C. (2008): Fuzzy Multi-Criteria Decision Making. Theory and Application
with Recent Development. Optimization and its Applications, Vol. 16, Springer,
New York.
Kahraman C., Buyukozkan G., Ates N.Y. (2007): A Two-Phase Multi-Attribute
Decision Making Approach for New Products Introduction. “Information
Sciences”, 177, pp. 1567-1582.
Liu S., Lin Y. (2006): Grey Information. Theory and Practical Applications. Advanced
Information and Knowledge Processing, Springer, London.
Łuczak A., Wysocki F. (2006): Rozmyta wielokryterialna metoda porządkowania
liniowego obiektów. Taksonomia 13, Prace Naukowe, Wydawnictwo Akademii
Ekonomicznej, Wrocław, pp. 148-157.
Łuczak A., Wysocki F. (2008): Wykorzystanie rozmytej metody TOPSIS opartej na
zbiorach a-poziomów do porządkowania liniowego obiektów. Taksonomia 15,
Prace Naukowe, Wydawnictwo Akademii Ekonomicznej, Wrocław, pp. 337-
345.
Olson D.L. (2004): Comparison of Weights in TOPSIS Models. “Mathematical and
Computer Modeling”.
Rigopoulos G., Psarras J., Askounis D. (2008): Group Decision Methodology for
Collaborative Multicriteria Assignment. “Word Applied Sciences Journal”, 4(1),
pp. 155-163.
Roszkowska E. (2009): Application the TOPSIS Methods for Ordering Offers in Buyer-
-Seller Transaction. Optimum − Studia Ekonomiczne, Vol. 3(43), pp. 117-133.
Saghafian S., Hejazi S.R. (2005): Multi-Criteria Group Decision Making Using
a Modified Fuzzy TOPSIS Procedure. Proceedings of the 2005 International
Conference on Computational Intelligence and Modeling, Control and Auto-
mation, and International Conference Intelligent Agents, Web Technologies
and Internet Commerce.
230 Ewa Roszkowska

Saaty T.L. (1980): The Analytic Hierarchy Process. McGraw Hill, New York.
Shih H.S., Shyur H.J., Lee E.S. (2007): An Extension of TOPSIS for Group Decision
Making. “Mathematical and Computer Modelling”, Vol. 45, pp. 801-813.
Zavadskas E.K., Turskis Z., Tamosaitiene J. (2008): Construction Risk Assessment
of Small Scale Objects by Applying the TOPSIS Method with Attributes Values
Determined at Intervals. The 8th International Conference “Reliability and
Statistic in Transportation and Communication”, Latvia.