DOI 10.1007/s00521-015-1891-2
ORIGINAL ARTICLE
Keywords Fuzzy set Intuitionistic fuzzy set Multiattribute group decision-making Neutrosophic set
Single-valued neutrosophic set TOPSIS
1 Introduction
Multiple attribute decision-making (MADM) problems
with quantitative or qualitative attribute values have broad
applications in the area of operation research, management
science, urban planning, natural science, and military affairs, etc. The attribute values of MADM problems cannot
be expressed always with crisp numbers because of ambiguity and complexity of attribute. In classical MADM
methods, such as technique for order preference by similarity to ideal solution (TOPSIS) developed by Hwang and
Yoon [1], PROMETHEE [2], VIKOR [3], ELECTRE [4],
the weight of each attribute and ratings of alternative are
presented by crisp numbers. However, in real world, decision maker may prefer to evaluate attributes by using
linguistic variables rather than exact values because of
partial knowledge about the attribute and lack of information processing capabilities of the problem domain. In
such situation, a preference information of alternatives
provided by the decision makers may be vague, imprecise,
or incomplete. Fuzzy set [5] introduced by Zadeh is one of
such tool that utilizes impreciseness in a mathematical
form. MADM problem with imprecise information can be
modeled quite well by using fuzzy set theory into the field
of decision-making. Chen [6] extended the TOPSIS
method for solving multi-criteria decision-making problems in fuzzy environment. However, fuzzy set can only
focus on the membership degree of vague parameters or
events. It fails to handle non-membership degree and indeterminacy degree of imprecise parameters. In 1986,
123
123
SVNS information. The information provided by different domain experts in MAGDM problems about alternative and attribute values takes the form of singlevalued neutrosophic set. In a group decision-making
process, neutrosophic weighted averaging operator
needs to be used to aggregate all the decision makers
opinions into a single opinion for rating the selected
alternatives.
The remaining part of this paper is organized as follows:
Sect. 2 briefly introduces some preliminaries relating to
neutrosophic set and the basics of single-valued neutrosophic set. In Sect. 3, basics of TOPSIS method are discussed. Section 4 is devoted to develop TOPSIS method
for MADM under simplified neutrosophic environment. In
Sect. 5, an illustrative example is provided to show the
effectiveness of the proposed approach. Finally, Sect. 6
presents the concluding remarks.
0 TN x IN x FN x 3 :
~ T ~ x; I ~ x; F ~ x and B~
Definition 5 Let A
A
A
A
TB~ x; IB~ x; FB~x be any two SVNSs, then Wang
et al. [11] defined the following set of operations as:
1.
2.
~ B~ if and only if T ~ x T ~ x; I ~ x
A
B
A
A
IB~ x; FA~ x FB~x for all x 2 X.
~ B~ if and only if A
~ B~ and B~ A
~ for all x 2 X.
A
~ c fxj F ~ x; 1 I ~ x; T ~ x jx 2 Xg for all x
3. A
A
A
A
2 X:
1
Single-valued neutrosophic set is a special case of neutrosophic set. It can be used in real scientific and engineering applications. In the following sections, some basic
definitions, operations, and properties regarding singlevalued neutrosophic sets [11] are provided.
4.
Definition 6
5.
for all
x2X
1.
~
B~ hT ~ x T ~ x T ~ x:T ~x;
A
B
B
A
A
IA~ x:IB~x; FA~ x:FB~ xi for all x 2 X:
2.
~ B~ hT ~ x:T ~ x; I ~ x I ~x
A
B
B
A
A
3.
~ [ B~ maxT ~ x; T ~x;
A
B
A
minIA~ x; IB~x; minFA~ x; FB~ xi for all x 2 X.
~ \ B~ minT ~ x; T ~ x;
A
4.
123
~ B
~
DEucl A;
v
)
u n (
uX TA~ xi TB~ xi 2 IA~ xi IB~ xi 2
t
2
FA~ xi FB~xi
i1
4
and the normalized Euclidean distance between two
~ and B~ can be defined as follows:
SVNSs A
~ B
~
DNEucl A;
v
(
2
2 )
u
n
u1 X
T
x
T
x
I
x
I
x
~
~
~
~
i
i
i
i
B
B
A
A
t
2
3n i1
FA~ xi FB~ xi
5
for
i 1; 2; . . .; m and j 1; 2; . . .; n:
6
Definition 8
PIS A v
1 ; v2 ; . . .vn ;
(
!
7
and
3 TOPSIS
TOPSIS method is used to determine the best alternative
from the concepts of the compromise solution. The best
compromise solution should have the shortest Euclidean
distance from the ideal solution and the farthest Euclidean
distance from the negative ideal solution. The procedures
of TOPSIS can be described as follows. Let A
fA1 ; A2 ; . . .Am g be the set of alternatives, C
fC1 ; C2 ; . . .Cn g be the set of criteria and D dij ,
i 1; 2; . . .; m, j 1; 2; . . .; n, be the performance ratings
with the criteria weight vector W fwj jj 1; 2; . . .; ng.
TOPSIS method is presented with these following steps.
3.1 Step 1. Normalization the decision matrix
The normalized value dijN is calculated as follows:
123
NIS A v
1 ; v2 ; . . .vn ;
(
!
8
where J1 and J2 are the benefit and cost-type criteria,
respectively.
3.4 Step 4. Calculation of the separation measures
for each alternative from the PIS and the NIS
The separation values for the PIS can be measured by using
the n-dimensional Euclidean distance, which is given as:
v
uX
2
u n
t
D
i 1; 2; . . .m:
v
v
9
ij
j
i
j1
2
u n
Di t
i 1; 2; . . .m:
vij v
j
j1
10
for
i 1; 2; . . .m:
11
D dij m
n A2
...
C1
0
d11
Bd
B 21
B
@ ...
C2
d12
...
...
d22
...
...
...
Cn
1
d1n
d2n C
C
C
... A
Am
d1m
d2m
...
dmn
A1
12
4.1 Step 1. Determination of the most important
attribute
Generally, there are many criteria or attributes in decisionmaking problems, where some of them are important and
others may not be so important. So it is crucial to select the
proper criteria or attributes for decision-making situation.
The most important attributes may be chosen with the help
of expert opinions or by some other method that are
technically sound.
4.2 Step 2. Construction of the decision matrix
with SVNSs
It is assumed that the rating of each alternative with respect
to each attribute is expressed as SVNS for MADM
problem. The neutrosophic values associated with the alternatives for MADM problems can be represented in the
following decision matrix:
D E
DN~ dijs
Tij ; Iij ; Fij m
n
13
m
n
A1
A2
...
Am
C1
hT11 ; I11 ; F11 i
B hT ; I ; F i
B 21 21 21
B
@
...
hTm1 ; Im1 ; Fm1 i
0
C2
hT12 ; I12 ; F12 i
hT22 ; I22 ; F22 i
...
hTm2 ; Im2 ; Fm2 i
...
...
...
...
...
Cn
1
hT1n ; I1n ; F1n i
hT2n ; I2n ; F2n i C
C
C
A
...
hTmn ; Imn ; Fmn i
14
In the matrix DN~ Tij ; Iij ; Fij m
n , Tij ,Iij and Fij denote
the degree of truth-membership value, indeterminacymembership value and falsity-membership value of alternative Ai with respect to attribute Cj satisfying the following properties under the single-valued neutrosophic
environment:
1.
2.
and
123
can be considered as the set of all ratings whose truthmembership value is zero.
V Tij ; Iij ; Fij
where
Tij 0; 0\Iij 1 and 0\Fij 1:
for i 1; 2; . . .; m and j 1; 2; . . .; n.
Consideration of V should be avoided in decisionmaking process.
Definition 11 (Tolerable neutrosophic ratings) Excluding the area of highly acceptable ratings and unacceptable
ratings from a neutrosophic cube, tolerable neutrosophic
rating area H D \ :K \ :C can be determined. The
tolerable neutrosophic rating Z considered with below
average truth-membership degree, above average indeterminacy degree and above average falsity-membership degree are taken in decision-making process. Z can be
defined by the following expression
Z Tij ; Iij ; Fij
where 0\Tij \0:5; 0:5\Iij \1 and 0:5\Fij \1:
for i 1; 2; . . .; m and j 1; 2; . . .; n.
Definition 12
D
Fuzzification of SVNS N~ fxj TN~ x;
1
q
f1 TN~ x2 IN~ x2 FN~ x2 g=3;
0;
for
8x 2 U [ Z
for
8x 2 V
15
4.3 Step 3. Determination of the weights of decision
makers
Let us assume that the group of p decision makers having
their own decision weights. Thus, the importance of the
decision makers in a committee may not be equal to each
other. Let us assume that the importance of each decision
maker is considered with linguistic variables and expressed
it by neutrosophic numbers.
Let Ek hTk ; Ik ; Fk i be a neutrosophic number defined
for the rating of kth decision maker. Then, according to
123
Pp
k1
wk 1
4.4 Step 4. Construction of the aggregated singlevalued neutrosophic decision matrix based
on decision makers assessments
k
Let Dk dij m
n be the single-valued neutrosophic decision matrix of the k th decision maker and W
T
w1 ; w2 . . .; wp be the weight vector of decision maker such
that each wk 2 0; 1. In the group decision-making process,
all the individual assessments need to be fused into a group
opinion to make an aggregated neutrosophic decision matrix.
This aggregated matrix can be obtained by using singlevalued neutrosophic weighted averaging (SVNWA) aggregation operator proposed by Ye [17] for SVNSs as follows:
D dij m
n where,
1 2
p
dij SVNSWAW dij ; dij ; . . .; dij
1
w1 dij
w2 dij
wp dij
*
+
p
p
p
Y
p wk
p wk
p wk
1
1 Tij
;
Iij
;
Fij
k1
k1
k1
17
Therefore, the aggregated neutrosophic decision matrix is
defined as follows:
D dij m
n Tij ; Iij ; Fij m
n
18
A1
A2
...
Am
0
B
B
B
@
C1
hT11 ; I11 ; F11 i
hT21 ; I21 ; F21 i
...
C2
hT12 ; I12 ; F12 i
hT22 ; I22 ; F22 i
...
...
...
...
...
Cn
hT1n ; I1n ; F1n i
hT2n ; I2n ; F2n i
...
...
1
C
C
C
A
19
where, dij Tij ; Iij ; Fij is the aggregated element of
neutrosophic decision matrix D for i 1; 2; . . .m and
j 1; 2; . . .n.
4.5 Step 5. Determination of the attribute weights
In the decision-making process, decision makers may feel
that all attributes are not equally important. Thus, every
wj ; wj . . .; wj be the neutrosophic number(NN) assigned to the attribute Cj by the kth decision maker. Then the
combined weight W fw1 ; w2 . . .; wn g of the attribute can
be determined by using SVNWA aggregation operator [17].
1
2
p
wj SVNWAW wj ; wj ; . . .; wj
1
w1 wj
w2 wj
wp wj
*
+
p
p
p
Y
p wk
p wk
p wk
1
1 Tj
;
Ij
;
Fj
k1
k1
k1
QN~ d1 ; d2 ; . . .; dnw
24
D
E
where, djw Tjw ; Ijw ; Fjw for j 1; 2; . . .; n.
(
!
!)
w
Tjw
maxfTij j gjj 2 J1 ;
i
20
21
W fw1 ; w2 ; . . .; wn g
where, wj Tj ; Ij ; Fj for j 1; 2; . . .n.
A1
A2
...
Am
C1
0 w1 w1 w1
T11 ; I11 ; F11
B T w1 ; I w1 ; F w1
B 21 21 21
B
@
...
w1 w1 w1
Tm1 ; Im1 ; Fm1
m n
C2
w2 w2
w2
T12
; I12 ; F12
w2 w2 w2
T22 ; I22 ; F22
...
...
...
...
...
...
w2 w2
w2
; Im2 ; Fm2
Tm2
Cn
wn wn
wn 1
; I1n ; F1n
T1n
wn wn wn C
T2n ; I2n ; F2n C
C
A
...
w w w
Tmnn ; Imnn ; Fmnn
!)
Fjw
In this section, the obtained weights of attribute and aggregated neutrosophic decision matrix need to be further
fused to make the aggregated weighted neutrosophic decision matrix.
The aggregated weighted neutrosophic decision matrix
can be obtained by using the multiplication formula (3) of
two neutrosophic sets as:
D E
D
E
w
w
w
w
D W Dw dij j
Tij j ; Iij j ; Fij j
22
m
n
25
(
Ijw
minfTij j gjj 2 J2
i
minfFij j gjj 2 J1 ;
i
maxfFij j gjj 2 J2
i
27
can be defined by
and Q
N~
Q
d1w ; d2w ; . . .; dnw
28
N~
D
E
where, djw Tjw ; Ijw ; Fjw for j 1; 2; . . .; n.
(
!
!)
w
Tjw
minfTij j gjj 2 J1 ;
i
maxfTij j gjj 2 J2
i
29
(
!
w
Ijw
maxfIij j gjj 2 J1 ;
i
!)
w
minfIij j gjj 2 J2
i
30
23
D
E
w
w
w
w
Here, dij j Tij j ; Iij j ; Fij j is an element of the aggregated
weighted neutrosophic decision matrix Dw for i
1; 2; . . .; m and j 1; 2; . . .; n.
Fjw
D E
Let DN~ dijw
m n
Tij ; Iij ; Fij m
n be a SVNS-based
decision matrix, where, Tij , Iij and Fij are the membership
degree, indeterminacy degree and non-membership degree
of evaluation for the attribute Cj with respect to the alternative Ai .
!
w
maxfFij j gjj 2 J1 ;
i
!)
w
minfFij j gjj 2 J2
i
31
123
wj
w
Di
Eu dij ; dj
v
8
2 9
u
u
> T wj xj T w xj I wj xj I w xj =
>
n <
u1 X
ij
ij
j
j
u
t3n
2
>
>
w
;
j1 :
Fij j xj Fjw xj
32
similarly, the normalized Euclidean distance measure of
D
E
w
w
w
each alternative
Tij j ; Iij j ; Fij j
from the RNNIS
D
E
Tjw ; Ijw ; Fjw can be written as:
wj
w
Di
Eu dij ; dj
v
8
2 9
u
wj
wj
u
w
w
>
>
<
=
n
T
x
T
x
I
x
I
x
u1 X
j
j
j
j
ij
ij
j
j
u
t3n
2>
>
w
j1 :
F j xj F w xj ;
ij
33
4.9 Step 9. Determination of the relative closeness
coefficient to the neutrosophic ideal solution
for SVNSs
The relative closeness coefficient of each alternative Ai
with respect to the neutrosophic positive ideal solution Q
N~
is defined as follows:
w
Di
dij j ; djw
Eu
Ci
34
wj
wj
w
w
Di
Di
Eu dij ; dj
Eu dij ; dj
where, 0 Ci 1.
4.10 Step 10. Ranking the alternatives
According to the relative closeness coefficient values larger
the values of Ci reflects the better alternative Ai for
i 1; 2; . . .; m.
p
1 0:01 0:01 0:01=3
p p p p
0:292
35
36
5 Numerical example
Let us suppose that a group of four decision makers
(DM1 ,DM2 , DM3 , DM4 ) intend to select the most suitable
tablet from the four initially chosen tablet A1 ; A2 ; A2 ; A4
by considering six attributes namely: Features C1 , Hardware C2 , Display C3 , Communication C4 , Affordable Price
C5 , Customer care C6 . Based on the proposed approach
discussed in Sect. 4, the considered problem is solved by
the following steps:
123
SVNNs
Good/important (G/I)
Bad/unimportant (B/UI)
Fair/medium (F/M)
DM-1
LT
e
W
DM-2
VI
Linguistic terms
SVNNs
Medium/fair (M/F)
Medium bad/medium law (MB/ML)
Very bad/low (VB/VL)
Bad/law (B/L)
37
F11 0:050:292
0:050:265
0:150:178
0:150:265
0:0813:
38
5.3 Step-3. Determine the weights of attribute
The linguistic terms shown in Table 1 are used to evaluate
each attribute. The importance of each attribute for every
Table 4 Assessments of
alternatives and attribute
weights given by four decision
makers
Good/high (G/H)
DM-4
DM-3
6
7
6
7
W 6 h0:765; 0:226; 0:182i; h0:692; 0:277; 0:251i; 7
4
5
h0:788; 0:200; 0:180i;
Alternatives Ai
Decision makers
C1
C2
C3
C4
C5
C6
A1
DM-1
VG
VG
DM-2
VG
VG
VG
DM-3
VG
VG
DM-4
DM-1
DM-2
MG
MG
DM-3
DM-4
DM-1
VG
VG
VG
VG
DM-2
VG
VG
VG
DM-3
VG
MG
MG
DM-4
VG
VG
MG
DM-1
VG
VG
DM-2
DM-3
DM-4
G
G
G
M
G
M
G
G
M
G
VG
VG
A2
A3
A4
Weights
DM-1
VI
VI
DM-2
VI
DM-3
DM-4
VI
VI
123
A1
A2
A3
A4
A1
A2
A3
A4
A1
A2
A3
A4
A1
A2
A3
A4
C1
h0:864; 0:136; 0:081i
B h0:667; 0:333; 0:277i
B
@ h0:880; 0:120; 0:067i
h0:667; 0:333; 0:277i
C2
h0:853; 0:147; 0:092i
h0:727; 0:273; 0:219i
h0:887; 0:113; 0:064i
h0:735; 0:265; 0:195i
C3
1
h0:800; 0:200; 0:150i
h0:667; 0:333; 0:277i C
C
h0:834; 0:166; 0:112i A
h0:768; 0:232; 0:180i
C4
h0:704; 0:296; 0:241i
B h0:744; 0:256; 0:204i
B
@ h0:779; 0:221; 0:170i
h0:727; 0:273; 0:221i
C5
h0:823; 0:177; 0:123i
h0:652; 0:348; 0:293i
h0:811; 0:189; 0:109i
h0:791; 0:209; 0:148i
C6
1
h0:864; 0:136; 0:081i
h0:608; 0:392; 0:336i C
C
h0:850; 0:150; 0:092i A
h0:808; 0:192; 0:127i
C1
h0:652; 0:328; 0:280i
B h0:504; 0:481; 0:434i
B
@ h0:664; 0:315; 0:269i
h0:504; 0:481; 0:434i
C2
h0:757; 0:243; 0:289i
h0:645; 0:355; 0:388i
h0:787; 0:213; 0:267i
h0:652; 0:348; 0:370i
C3
1
h0:612; 0:381; 0:305i
h0:510; 0:484; 0:409i C
C
h0:638; 0:354; 0:274i A
h0:588; 0:406; 0:329i
C4
h0:487; 0:491; 0:432i
B h0:515; 0:402; 0:404i
B
@ h0:539; 0:437; 0:378i
h0:503; 0:474; 0:417i
C5
h0:649; 0:342; 0:281i
h0:514; 0:478; 0:420i
h0:639; 0:351; 0:269i
h0:623; 0:367; 0:301i
C6
1
h0:605; 0:371; 0:305i
h0:426; 0:557; 0:498i C
C
h0:595; 0:381; 0:314i A
h0:566; 0:412; 0:340i
3
h0:504; 0:481; 0:434i; h0:645; 0:355; 0:388i;
6
7
4 h0:510; 0:484; 0:409i; h0:487; 0:491; 0:432i; 5
Q
N~
h0:514; 0:478; 0:420i; h0:426; 0:557; 0:498i
44
w
w
is calculated as
123
is calculated as
Di
Eucl
Di
Eucl
Ci
0.8190
(42)
41
5.5 Step-5. Determination of the neutrosophic
relative positive ideal solution
and the neutrosophic relative negative ideal
solution
(39)
A1
0.0283
0.1281
A2
0.3472
0.0490
0.1158
A3
0.0224
0.1382
0.8605
A4
0.0900
0.0831
0.4801
6 Conclusions
This paper is devoted to present a new TOPSIS-based
approach for MAGDM under simplified neutrosophic environment. In the evaluation process, the ratings of each
alternative with respect to each attribute are given as linguistic variables characterized by single-valued neutrosophic numbers. Neutrosophic aggregation operator is used
to aggregate all the opinions of decision makers. Neutrosophic positive ideal and neutrosophic negative ideal solution are defined from aggregated weighted decision
matrix. Euclidean distance measure is used to determine
the distances of each alternative from positive as well as
negative ideal solutions for relative closeness coefficient of
each alternative. However, the author hopes that the concept presented in this paper may open up new avenue of
research in competitive neutrosophic decision-making
arena. TOPSIS method with neutrosophic set information
has enormous chance of success for multi-attribute decision-making problems. In future, the proposed approach
can be used for dealing with decision-making problems
such as personal selection in academia, project evaluation,
supplier selection, manufacturing systems, and many other
areas of management systems.
References
1. Hwang CL, Yoon K (1981) Multiple attribute decision making:
methods and applications. Springer, New York
2. Brans JP, Vinvke P, Mareschal B (1986) How to select and how
to rank projects: the PROMETHEE method. Eur J Oper Res
24:228238
3. Opricovic S, Tzeng GH (2004) Compromise solution by MCDM
methods: a comparative analysis of VIKOR and TOPSIS. Eur J
Oper Res 156:445455
123