Anda di halaman 1dari 6

Environ Monit Assess (2010) 168:339344

DOI 10.1007/s10661-009-1117-0
A modied VIKOR method for multiple criteria analysis
Chia-Ling Chang
Received: 18 April 2009 / Accepted: 27 July 2009 / Published online: 12 August 2009
Springer Science + Business Media B.V. 2009
Abstract The VIKOR method was developed to
solve multiple criteria decision making (MCDM)
problems with conicting or non-commensurable
criteria. This method assumes that compromis-
ing is acceptable for conicting resolution. Al-
though the VIKOR method is a popular method
applied in multi-criteria analysis (MCA), it has
some problems when solving MCDM problems.
This study discussed existing problems in the tra-
ditional VIKOR method. The objective of this
study was to develop a modied VIKOR method
to avoid numerical difculties in solving problems
by the traditional VIKOR method. Several syn-
thetic experiments were designed and assessed to
verify the improvement of solution efciency of
the modied VIKOR method in MCA.
Keywords Multiple criteria analysis
Multiple criteria decision making (MCDM)
VIKOR
C.-L. Chang (B)
Department of Water Resources Engineering
and Conservation, Feng Chia University,
No. 100, Wenhwa Road, Seatwen, Taichung,
Taiwan 40724, Republic of China
e-mail: f89541201@ntu.edu.tw
Introduction
Multiple criteria decision making (MCDM)
method is a useful tool for determining the best
solution among several alternatives according
to multiple criteria or attributes with different
effects. Because criteria in multi-criteria analysis
(MCA) are usually in conict, the solution is
usually impossible to satisfy all criteria. Therefore,
Pareto optimality was addressed for applying in
MCA. If one criterion is improved, another one
or more criteria are worsened; it can be regarded
as a Pareto optimal solution (Pareto 1896; Kuhn
and Tucker 1951; Zadeh 1963). The development
of the compromise-ranking methods, e.g., VIKOR
and TOPSIS, is based on the concept of Pareto
optimality.
The VIKOR method was developed as a
MCDM method to solve problems with conict-
ing or non-commensurable criteria (Opricovic and
Tzeng 2004). It assumes that compromising is ac-
ceptable for conicting resolution. The VIKOR
method has been popularly applied in MCA
(Opricovic and Tzeng 2002, 2003). However, some
problems exist when solving multiple criteria op-
timization problems. To improve the value of the
VIKOR method, the objective of this study was to
improve the numerical procedure and to eliminate
some difculties that could appear in solving some
particular problems by the traditional VIKOR
method. The modied VIKOR method is novel,
340 Environ Monit Assess (2010) 168:339344
rst introduced in public academic eld and useful
for applying in environmental management.
Methodology
MCDM and the VIKOR method
MCDM method can be applied to evaluate the
alternatives according to multiple and conict-
ing criteria. Quantitative values and weights for
criteria are required for assessing the effect and
relative importance of criteria. The main steps
of MCDM includes: (1) developing alternatives;
(2) establishing evaluation criteria; (3) quantifying
the value and weights of criteria of each alterna-
tive; (4) assessing the alternatives and determining
a nal optimal or compromise solution (Opricovic
and Tzeng 2004). Since conicting criteria usu-
ally exist in MCDM problems, absolute optimal
solution is not easy to be determined. Thus, the
concept of compromise solution is signicant in
MCA.
The compromise-ranking method, VIKOR, is
an applicable technique for MCA. It was devel-
oped for multi-criteria optimization in complex
systems. The VIKOR method focuses on ranking
and selecting from the alternatives with conict-
ing criteria (Tzeng et al. 2005). Assuming that
each alternative is evaluated according to mul-
tiple criterion functions, the compromise rank-
ing is performed by comparing the measure of
closeness to the ideal alternative (Opricovic 1998;
Tzeng et al. 2002; Opricovic and Tzeng 2004). The
compromise-ranking algorithm of the traditional
VIKOR has following steps.
(a) Step 1: The various alternatives are denoted
as x
1
, x
2
,. . . . . . , x
m
. For an alternative x
j
,
the merit of the ith aspect is denoted by f
ij
,
i.e., f
ij
is the value of ith criterion function
for the alternative x
j
. m is the number of
alternatives; n is the number of criteria.
(b) Step 2: Determine the maximum f

i
and the
minimum f

i
values of all criterion functions,
i = 1. . . . . . n.
f

i
= j max f
ij
=max

f
ij

| j = 1, 2, ......, m

(1)
f

i
= j min f
ij
=min

f
ij

| j = 1, 2, ......, m

(2)
(c) Step 3: Compute the values S
j
and R
j
, j =
1. . . . . . m.
S
j
=
n

i=1
w
i

i
f
ij

i
f

(3)
R
j
= i max

w
i

i
f
ij

i
f

| i
= 1, 2, . . . . . . , n

(4)
where S
j
and R
j
represent the utility mea-
sure and the regret measure, respectively,
for the alternative x
j
. w
i
are the weight of
ith criterion, which represents the relative
importance of criterion.
(d) Step 4: Compute the values Q
j
, j = 1. . . . . . m.
Q
j
= v(S
j
S

)/(S

)
+ (1 v)(R
j
R

)/(R

) (5)
S

= j min S
j
= min

S
j

| j = 1, 2, . . . . . . , m

(6)
S

= j max S
j
= max

S
j

| j = 1, 2, . . . . . . , m

(7)
R

= j min R
j
= min

R
j

| j = 1, 2, . . . . . . , m

(8)
R

= j max R
j
= max

R
j

| j = 1, 2, . . . . . . , m

(9)
where v is the weight for the strategy of max-
imum group utility and 1 v is the weight of
the individual regret. v is usually set to 0.5
(Kackar 1985; Opricovic 1998).
(e) Step 5: Rank the alternatives by Q
j
. The
less the value of Q
j
is, the better decision
of the alternatives is. Since it provides a
Environ Monit Assess (2010) 168:339344 341
maximum group utility of the majority and
a minimum individual regret of the oppo-
nent, the obtained compromise solution is
acceptable by decision makers (Opricovic
and Tzeng 2007; Tong et al. 2007).
Modied VIKOR method
Although the VIKOR method is an effective tool
for MCDM problems, some errors would occur in
multi-criteria optimization calculation. The prob-
lems existing in the traditional VIKOR method
are listed in Table 1. When the ith criterion
has no difference among alternatives, i.e., f
i1
=
f
i2
= ... = f
im
= f

i
= f

i
, error calculation will
occur and S
j
and R
j
cannot be determined by
Eqs. (3) and (4). When the utility measure or/and
regret measure has no difference among alterna-
tives, the traditional VIKOR method cannot de-
termine Q
j
by Eq. (5). To overcome the numerical
problems existing in Eqs. (3) and (4), selected
criteria should be meaningful. If the criterion has
no difference among alternatives, it is not suitable
for regarding as a measure. This study developed
a modied VIKOR method having logic judg-
ment. The modied VIKOR index Q

j
is dened
as Eq. (10). Through the feasible logic judgment,
the alternatives can be ranked by the modied
VIKORindex and the problems existing in Eq. (5)
can be solved.
Table 1 Existing
problems in the
traditional VIKOR
method and solutions for
the problems
Q

j
is the modied
VIKOR index
Problems Occurrence reasons Solutions
f

i
= f

i
Some criteria have no Avoid meaningless

i
= max f
ij
; f

i
= min f
ij

difference among the criteria


alternatives
S

= S

The utility measure has If S

= S

= max S
j
; S

= min S
j

no difference among Q

j
=

R
j
R

the alternatives
R

= R

The regret measure has If R

= R

= max R
j
; R

= min R
j

no difference among Q

j
=

S
j
S

the alternatives
Q

j
=

S
j
S

+(1 v)

R
j
R

= S

= R

R
j
R

= S

= R

S
j
S

= R

= S
k = constant S

= S

= R

(10)
Synthetic case design
Synthetic data can replace measured data to
simplify complex systems (Hromadka 1996a, b).
Proper synthetic case design is useful for under-
standing the characteristics and advantages of an
original methodology. This study designed three
synthetic experiments for discussing existing prob-
lems in the traditional VIKOR method. In the
synthetic experiments, there were three alterna-
tives (x
1
, x
2
, and x
3
) and ve criteria (c
1
, c
2
, c
3
, c
4
,
and c
5
). The weights of criteria had an equal value,
i.e., 0.2. The value of v was given by 0.5. Table 2
shows the values of criteria for each alternative.
This study compared the calculation to identify
the priority-ranking of alternatives by the tradi-
tional and modied VIKOR method according to
the synthetic experimental data.
342 Environ Monit Assess (2010) 168:339344
Table 2 Synthetic experimental data
Alternatives\criteria c
1
c
2
c
3
c
4
c
4
Experiment 1
x
1
1 3 1 10 5
x
2
1 10 1 7 10
x
3
1 1 1 1 1
Experiment 2
x
1
10 4 10 1 4
x
2
3 10 1 10 1
x
3
1 1 7 9 10
Experiment 3
x
1
10 5 10 1 1
x
2
5 10 1 10 1
x
3
1 1 5 10 10
Results and discussion
Problems in the traditional VIKOR method
In the synthetic experiment 1, some selected crite-
ria are meaningless. Criterion 1 (c
1
) and criterion
3 (c
3
) has no difference among these alternatives
(x
1
, x
2
, and x
3
), i.e., f

1
= f

1
= f
11
= f
12
= f
13
=
1 and f

3
= f

3
= f
31
= f
32
= f
33
= 1. Due to zero
denominators, i.e., f

3
f

1
= f

3
f

3
= 0, S
j
and R
j
cannot be calculated by Eqs. (3) and (4).
Moreover, the alternatives have different utility
measure and the same regret measure in the syn-
thetic experiment 2. The utility measure and re-
gret measure both have no difference among the
alternatives in the synthetic experiment 3. Thus,
the value of Q
j
is unavailable even though S
j
and
R
j
can be determined in experiments 2 and 3.
Table 3 lists the value of S
j
and R
j
for the alter-
natives in these synthetic experiments. It shows
that the priority-ranking of alternatives cannot
be identied by the traditional VIKOR method,
since the Q
j
index is invalid.
Table 4 Priority-ranking of alternatives by the modied
VIKOR method
Experiment\alternatives x
1
x
2
x
3
1 2 1 3
2 1 3 2
3
represents that the alternatives have the same priority-
ranking according to the utility measure and regret
measure
Improvement of the modied VIKOR method
Table 4 shows the priority-ranking of alternatives
in these synthetic cases by the modied VIKOR
method. In experiment 1, meaningless criteria (c
1
and c
3
) should be excluded. The weights for the
remaining criteria were given by 1/3. When the
value of selected criteria are different among the
alternatives, S
j
and R
j
can be determined by
Eqs. (3) and (4) and the VIKOR index Q
j
can
be computed by Eq. (5). The priority-ranking of
alternatives can be identied. In experiment 1,
alternative 2 (x
2
) is the optimal selection which
has maximumutility measure and minimumregret
measure.
In experiments 2 and 3, the traditional VIKOR
method cannot be successfully applied to calculate
the VIKOR index Q
j
. Therefore, the priority-
ranking of alternatives cannot be identied by
the traditional VIKOR method but the modied
VIKOR method with logic judgment. In experi-
ment 2, the modied VIKOR indices Q

1
, Q

2
, and
Q

3
are 0, 1, and 0.25 for the alternatives 1, 2, and
3 (x
1
, x
2
, and x
3
) respectively. According to the
modied VIKOR index Q

j
, alternative 1 (x
1
) has
the priority in experiment 2, since the modied
VIKOR index Q

1
is the least one among Q

j
( j =
1 3). The alternatives have the same value of
Table 3 Utility measure and regret measure for the alternatives in these experiments
Experiment\index S
1
S
2
S
3
S* S

R
1
R
2
R
3
R* R

1
2 0.47 0.56 0.49 0.47 0.56 0.20 0.20 0.20 0.20 0.20
3 0.51 0.51 0.51 0.51 0.51 0.20 0.20 0.20 0.20 0.20
Environ Monit Assess (2010) 168:339344 343
Table 5 Difference
between the traditional
VIKOR method and the
modied VIKOR method
Comparison item The traditional VIKOR The modied VIKOR
method method
Logic judgment
The alternatives have the Invalid index Q
j
Modied VIKOR
same utility measure indexQ

j
can identify
the priority-ranking
of alternatives
The alternatives have the Invalid index Q
j
Modied VIKOR
same regret measure indexQ

j
can identify
the priority-ranking
of alternatives
Both the utility measure Invalid index Q
j
The alternatives have the same
and regret measure priority-ranking according
are the same in the to utility measure and
alternatives regret measure
the modied VIKOR index Q

j
, i.e., Q

1
= Q

2
=
Q

3
, in experiment 3. Thus, the alternatives have
the same priority-ranking according to the equal
utility measure and regret measure.
Table 5 compares the traditional VIKOR
method and the modied VIKOR method. The
improvement of the modied VIKOR method can
be clearly understood. When there are only a
few alternatives in MCDM problems or the char-
acteristics of the alternatives are similar to one
another, the utility measure or/and regret measure
would have high possibility to be the same in
each alternative. When the alternatives have the
same utility measure and different regret measure
or when they have the same regret measure and
different utility measure, the modied VIKOR
method having logic judgment can be applied to
solve the problems resulted from zero denomina-
tors in the calculation by the traditional VIKOR
method. When both the utility measure and regret
measure are the same in the alternatives, the mod-
ied VIKOR method cannot identify the priority-
ranking of alternatives. However, invalid index
and false calculation, which would occur in the
traditional VIKOR method, can be avoided by the
modied VIKOR method.
Application in environmental management
The VIKOR method has been applied in various
elds. However, existing problems in the tradi-
tional VIKOR method is a restriction for solving
MCDM problems. The modied VIKOR method
developed in this study can overcome invalid so-
lution problems. Therefore, it can be much use-
ful for applying in environmental management.
Environmental vulnerability analysis, classied
land-use management, resource distribution, and
environmental conservation strategies can depend
on the MCA by the modied VIKOR method.
When the alternatives have equal utility mea-
sure or regret measure, the logic judgment in the
modied VIKOR method can avoid numerical
difculties for solving multi-criteria optimization
problems.
Conclusions
The compromise solution by VIKOR is with high
acceptance, because the priority-ranking of alter-
native by VIKOR is the closest to the optimal
solution. However, the calculation by the tradi-
tional VIKOR method would bring fault and the
VIKOR index Q
j
would be unavailable. These
problems lower the value of the compromise
ranking method. This study developed a modi-
ed VIKOR method to solve existing problems
in the traditional VIKOR method. Because the
modied VIKOR method has logic judgment, it
can avoid numerical difculties resulted from zero
denominator. The modied VIKOR method still
344 Environ Monit Assess (2010) 168:339344
compares the alternatives according to the utility
measure and regret measure. Thus, when the al-
ternatives both have the same utility measure and
regret measure, they will have the same priority-
ranking. Since the modied VIKOR index Q

j
is valid index in any situations, any ranking re-
sults can be reasonably explained by the modied
VIKOR method.
References
Hromadka, T. V. (1996a). A rainfall-runoff probabilistic
simulation program: 1. Synthetic data generation. En-
vironmental Software, 11(4), 235242.
Hromadka, T. V. (1996b). A rainfall-runoff probabilistic
simulation program: 2. Synthetic data analysis. Envi-
ronmental Software, 11(4), 243249.
Kackar, R. N. (1985). Off-line quality control, parameter
design and the Taguchi method. Journal of Quality
Technology, 17, 176188.
Kuhn, H. W., & Tucker, A. W. (1951). Nonlinear pro-
gramming. In J. Neyman (Ed.), Proceedings of the
second Berkeley symposium on mathematical statistics
and probability (pp. 481492). Berkeley: University of
California Press.
Opricovic, S. (1998). Multi-criteria optimization of civil
engineering systems. Belgrade: Faculty of Civil
Engineering.
Opricovic, S., & Tzeng, G. H. (2002). Multicriteria plan-
ning of post-earthquake sustainable reconstruction.
Computer-Aided Civil and Infrastructure Engineering,
17(3), 211220.
Opricovic, S., & Tzeng, G. H. (2003). Fuzzy multicriteria
model for post-earthquake land-use planning. Natural
Hazards Review, 4(2), 5964.
Opricovic, S., & Tzeng, G. H. (2004). Compromise solu-
tion by MCDM methods: A comparative analysis of
VIKOR and TOPSIS. European Journal of Opera-
tional Research, 156, 445455.
Opricovic, S., & Tzeng, G. H. (2007). Extended VIKOR
method in comparison with outranking methods.
European Journal of Operational Research, 178,
514529.
Pareto, V. (1896). Cours dEconomie Politique. Geneva:
Droz.
Tong, L. I., Chen, C. C., & Wang, C. H. (2007). Optimiza-
tion of multi-response processes using the VIKOR
method. International Journal of Advanced Manufac-
turing Technology, 31, 10491057.
Tzeng, G. H., Teng, M. H., Chen, J. J., & Opricovic, S.
(2002). Multicriteria selection for a restaurant location
in Taipei. International Journal of Hospitality Manage-
ment, 21(2), 171187.
Tzeng, G. H., Lin, C. W., & Opricovic, S. (2005).
Multi-criteria analysis of alternative-fuel buses for
public transportation. Energy Policy, 33, 1373
1383.
Zadeh, L. A. (1963). Optimality and non-scalar-valued per-
formance criteria. IEEE Transactions on Automatic
Control, 8(1), 5960.

Anda mungkin juga menyukai