Neural Comput & Applic (2016) 27:727–737

DOI 10.1007/s00521-015-1891-2

ORIGINAL ARTICLE

TOPSIS method for multi-attribute group decision-making

under single-valued neutrosophic environment
Pranab Biswas1 • Surapati Pramanik2 • Bibhas C. Giri1

Received: 17 November 2014 / Accepted: 18 March 2015 / Published online: 8 April 2015
 The Natural Computing Applications Forum 2015

Abstract A single-valued neutrosophic set is a special Keywords Fuzzy set  Intuitionistic fuzzy set  Multi-
case of neutrosophic set. It has been proposed as a gener- attribute group decision-making  Neutrosophic set 
alization of crisp sets, fuzzy sets, and intuitionistic fuzzy Single-valued neutrosophic set  TOPSIS
sets in order to deal with incomplete information. In this
paper, a new approach for multi-attribute group decision-
making problems is proposed by extending the technique 1 Introduction
for order preference by similarity to ideal solution to sin-
gle-valued neutrosophic environment. Ratings of alterna- Multiple attribute decision-making (MADM) problems
tive with respect to each attribute are considered as single- with quantitative or qualitative attribute values have broad
valued neutrosophic set that reflect the decision makers’ applications in the area of operation research, management
opinion based on the provided information. Neutrosophic science, urban planning, natural science, and military af-
set characterized by three independent degrees namely fairs, etc. The attribute values of MADM problems cannot
truth-membership degree (T), indeterminacy-membership be expressed always with crisp numbers because of am-
degree (I), and falsity-membership degree (F) is more ca- biguity and complexity of attribute. In classical MADM
pable to catch up incomplete information. Single-valued methods, such as technique for order preference by simi-
neutrosophic set-based weighted averaging operator is used larity to ideal solution (TOPSIS) developed by Hwang and
to aggregate all the individual decision maker’s opinion Yoon [1], PROMETHEE [2], VIKOR [3], ELECTRE [4],
into one common opinion for rating the importance of the weight of each attribute and ratings of alternative are
criteria and alternatives. Finally, an illustrative example is presented by crisp numbers. However, in real world, de-
provided in order to demonstrate its applicability and ef- cision maker may prefer to evaluate attributes by using
fectiveness of the proposed approach. linguistic variables rather than exact values because of
partial knowledge about the attribute and lack of infor-
mation processing capabilities of the problem domain. In
such situation, a preference information of alternatives
provided by the decision makers may be vague, imprecise,
& Pranab Biswas or incomplete. Fuzzy set [5] introduced by Zadeh is one of
[email protected]
such tool that utilizes impreciseness in a mathematical
Surapati Pramanik form. MADM problem with imprecise information can be
[email protected]
modeled quite well by using fuzzy set theory into the field
Bibhas C. Giri of decision-making. Chen [6] extended the TOPSIS
[email protected]
method for solving multi-criteria decision-making prob-
1
Department of Mathematics, Jadavpur University, lems in fuzzy environment. However, fuzzy set can only
Kolkata 700032, West Bengal, India focus on the membership degree of vague parameters or
2
Department of Mathematics, Nandalal Ghosh B.T College, events. It fails to handle non-membership degree and in-
Panpur, Narayanpur 743126, West Bengal, India determinacy degree of imprecise parameters. In 1986,

123
728 Neural Comput & Applic (2016) 27:727–737

Atanassov [7] introduced intuitionistic fuzzy set (IFS) SVNS information. The information provided by dif-
characterized by membership and non-membership degrees ferent domain experts in MAGDM problems about al-
simultaneously. ternative and attribute values takes the form of single-
Boran et al. [8] extended the TOPSIS method for multi- valued neutrosophic set. In a group decision-making
criteria intuitionistic decision-making problem. Pramanik process, neutrosophic weighted averaging operator
and Mukhopadhyaya [9] studied teacher selection in intu- needs to be used to aggregate all the decision makers’
itionistic fuzzy environment. However, in IFSs, sum of opinions into a single opinion for rating the selected
membership degree and non-membership degree of a alternatives.
vague parameter is less than unity. Therefore, a certain The remaining part of this paper is organized as follows:
amount of incomplete information or indeterminacy arises Sect. 2 briefly introduces some preliminaries relating to
in an intuitionistic fuzzy set. It cannot handle all types of neutrosophic set and the basics of single-valued neutro-
uncertainties successfully in different real physical prob- sophic set. In Sect. 3, basics of TOPSIS method are dis-
lems such as problems involving indeterminate cussed. Section 4 is devoted to develop TOPSIS method
information. for MADM under simplified neutrosophic environment. In
Smarandache [10] first introduced the concept of neu- Sect. 5, an illustrative example is provided to show the
trosophic set (NS) from philosophical point of view to effectiveness of the proposed approach. Finally, Sect. 6
handle indeterminate or inconsistent information that usu- presents the concluding remarks.
ally exists in real situation. A neutrosophic set is charac-
terized by a truth-membership degree, an indeterminacy-
membership degree, and a falsity-membership degree in- 2 Preliminaries of neutrosophic sets
dependently. An important feature of NS is that every and single-valued neutrosophic sets
element of the universe has not only a certain degree of
truth (T), but also a falsity degree (F) and indeterminacy In this section, some basic definitions of neutrosophic set
degree (I). This set is a generalization of crisp set, fuzzy defined by Smarandache [10] have been provided to de-
set, interval-valued fuzzy set, intuitionistic fuzzy set, in- velop the paper.
terval-valued intuitionistic fuzzy set, etc. However, NS is
difficult to apply directly in real engineering and scientific 2.1 Neutrosophic set
applications. In order to deal with difficulties, Wang
et al. [11] introduced a subclass of NS called single-valued Neutrosophic set is originated from neutrosophy, a new
neutrosophic set (SVNS) characterized by truth-member- branch of philosophy which reflects the origin, nature, and
ship degree, an indeterminacy-membership degree and a scope of neutralities, as well as their interactions with
falsity-membership degree. SVNS can be applied quite different ideational spectra [10].
well in real scientific and engineering fields to handle the
Definition 1 Let X be a universal space of points
uncertainty, imprecise, incomplete, and inconsistent infor-
(objects), with a generic element of X denoted by x. A
mation. Ye [12] studied multi-criteria decision-making
neutrosophic set N  X is characterized by a truth-
problem by using the weighted correlation coefficient of
membership function TN ðxÞ, an indeterminacy-member-
SVNSs. Ye [13] also developed single-valued neutrosophic
ship function IN ðxÞ and a falsity-membership function
cross-entropy for multi-criteria decision-making problems.
FN ðxÞ. TN ðxÞ, IN ðxÞ and FN ðxÞ are real standard or
Biswas et al. [14] proposed an entropy-based gray rela-
nonstandard subsets of ½ 0; 1þ , so that all three neutro-
tional analysis method for solving a multi- attribute deci-
sophic components TN ðxÞ ! ½ 0; 1þ , IN ðxÞ ! ½ 0; 1þ 
sion-making problem under SVNSs. Biswas et al. [15] also
and FN ðxÞ ! ½ 0; 1þ .
developed a new methodology for solving SVNS-based
MADM with unknown weight information. Zhang The set IN ðxÞ may represent not only indeterminacy, but
et al. [16] studied multi-criteria decision-making problems also vagueness, uncertainty, imprecision, error, contradic-
under interval neutrosophic set information. Ye [17] fur- tion, undefined, unknown, incompleteness, redundancy,
ther discussed multi-criteria decision-making problem by etc. [19, 20]. In order to catch up vague information, an
using aggregation operators for simplified neutrosophic indeterminacy-membership degree can be split into sub-
sets. Chi and Liu [18] discussed an extended TOPSIS components, such as ‘‘contradiction,’’ ‘‘uncertainty’’, and
method for interval neutrosophic set-based MADM ‘‘unknown’’. [21].
problems. The sum of three independent membership degrees
The objective of this paper was to extend the con- TN ðxÞ, IN ðxÞ and FN ðxÞ are related as follows [11]
cept of TOPSIS method for multi-attribute group de- 
0  TN ðxÞ þ IN ðxÞ þ FN ðxÞ  3þ :
cision-making (MAGDM) problems into MAGDM with

123
Neural Comput & Applic (2016) 27:727–737 729

 
Definition 2 The complement of neutrosophic set A is Definition 5 Let A ~ ¼ T ~ ðxÞ; I ~ ðxÞ; F ~ ðxÞ and B~ ¼
  A A A
denoted by Ac and is defined as TAc ðxÞ ¼ 1þ  TA ðxÞ, TB~ ðxÞ; IB~ ðxÞ; FB~ðxÞ be any two SVNSs, then Wang
c
IA ðxÞ ¼ 1þ  IA ðxÞ, and FA
c
ðxÞ ¼ 1þ  FA ðxÞ for all x 2 X et al. [11] defined the following set of operations as:
Definition 3 A neutrosophic set A is contained in other 1. A~  B~ if and only if T ~ ðxÞ  T ~ ðxÞ; I ~ ðxÞ
A B A
neutrosophic set B, i.e., A  B if and only if inf TA ðxÞ  IB~ ðxÞ; FA~ ðxÞ FB~ðxÞ for all x 2 X.
inf TB ðxÞ, sup TA ðxÞ  sup TB ðxÞ, inf IA ðxÞ inf IB ðxÞ, ~ ¼ B~ if and only if A ~  B~ and B~  A
~ for all x 2 X.
2. A
sup IA ðxÞ sup IB ðxÞ, inf FA ðxÞ inf FB ðxÞ, sup FA ðxÞ
sup FB ðxÞ, for all x in X.  
~ c ¼ fðxj F ~ ðxÞ; 1  I ~ ðxÞ; T ~ ðxÞ Þjx 2 Xg for all x
3. A A A A
2 X:
2.2 Single-valued neutrosophic set
ð1Þ
Single-valued neutrosophic set is a special case of neu- 
4. ~ [ B~ ¼ maxðT ~ ðxÞ; T ~ðxÞÞ;
A A B
trosophic set. It can be used in real scientific and engi-
minðIA~ ðxÞ; IB~ðxÞÞ; minðFA~ ðxÞ; FB~ ðxÞÞi for all x 2 X.
neering applications. In the following sections, some basic 
5. ~ \ B~ ¼ minðT ~ ðxÞ; T ~ ðxÞÞ;
A
definitions, operations, and properties regarding single- A B
valued neutrosophic sets [11] are provided. maxðIA~ ðxÞ; IB~ðxÞÞ; maxðFA~ ðxÞ; FB~ðxÞÞi for all x 2 X.

Definition 4 Let X be a universal space of points (objects), Liu and Wang defined the following set of operations for
with a generic element of X denoted by x. A single-valued SVNSs in [22] as:
neutrosophic set (SVNS) N~  X is characterized by a truth-
membership function TN~ ðxÞ, an indeterminacy-membership Definition 6 ~ and B~ be two SVNSs, then
Let A
function IN~ ðxÞ, and a falsity-membership function FN~ ðxÞ
with TN~ ðxÞ, IN~ ðxÞ, FN~ ðxÞ 2 ½0; 1 for all x 2 X. 1. ~
B~ ¼ hT ~ ðxÞ þ T ~ ðxÞ  T ~ ðxÞ:T ~ðxÞ;
A A B A B ð2Þ
IA~ ðxÞ:IB~ðxÞ; FA~ ðxÞ:FB~ ðxÞi for all x 2 X:
The sum of three membership functions of a SVNS N~ ,
the relation
2. ~  B~ ¼ hT ~ ðxÞ:T ~ ðxÞ; I ~ ðxÞ þ I ~ðxÞ
A
0  TN~ ðxÞ þ IN~ ðxÞ þ FN~ ðxÞ  3 for all x2X A B A B
 IA~ ðxÞ:IB~ðxÞ; FA~ ðxÞ þ FB~ ðxÞ ð3Þ
holds good. When X is continuous, a SVNS N~ can be  FA~ ðxÞ:FB~ðxÞi for all x 2 X:
written as
Z D 
E 3. ~ [ B~ ¼ maxðT ~ ðxÞ; T ~ðxÞÞ;
A A B
N~ ¼ TN~ ðxÞ; IN~ ðxÞ; FN~ ðxÞ jx; for all x 2 X: minðIA~ ðxÞ; IB~ðxÞÞ; minðFA~ ðxÞ; FB~ ðxÞÞi for all x 2 X.
x 
4. ~ \ B~ ¼ minðT ~ ðxÞ; T ~ ðxÞÞ;
A A B
When X is discrete, a SVNS N~ can be written as maxðIA~ ðxÞ; IB~ðxÞÞ; maxðFA~ ðxÞ; FB~ðxÞÞi for all x 2 X.
XD E
N~ ¼ TN~ ðxÞ; IN~ ðxÞ; FN~ ðxÞ jx; for all x 2 X:
x
2.3 Distance between two SVNSs
SVNS can be represented with the notation N~ ¼
D E
fðxj TN~ ðxÞ; IN~ ðxÞ; FN~ ðxÞ Þjx 2 Xg: Majumdar and Samanta [23] studied similarity and entropy
measure by incorporating euclidean distances of neutro-
Thus, finite SVNS can be presented by the ordered sophic sets.
tetrads:
Definition 7 (Euclidean distance) Let A ~ ¼ fðx1 jhT ~ ðx1 Þ;
N~ ¼ fðx1 jhTN~ ðx1 Þ; IN~ ðx1 Þ; FN~ ðx1 ÞiÞ; . . .; ðxn jhTN~ ðxn Þ;  A
IN~ ðxn Þ; FN~ ðxn ÞiÞg for all xi 2 Xði ¼ 1; 2; . . .; nÞ. For con- IA~ ðx1 Þ; FA~ ðx1 Þi; . . .; ðxn j TA~ ðxn Þ; IA~ ðxn Þ; FA~ ðxn Þ g and
D E   
venience, a SVNS N~ ¼ fðxj TN~ ðxÞ; IN~ ðxÞ; FN~ ðxÞ Þjx 2 B~ ¼ fðx1 j T ~ ðx1 Þ; I ~ðx1 Þ; F ~ðx1 Þ ; . . .; ðxn j T ~ ðxn Þ; I ~ðxn Þ;
B B B B B
D FB~ ðxn Þig be two SVNSs for xi 2 X ði ¼ 1; 2; . . .; nÞ: Then
Xg is denoted by the simplified symbol N~ ¼ TN~ ðxÞ; the Euclidean distance between two SVNSs A ~ and B~ can
IN~ ðxÞ; FN~ ðxÞi for all x 2 X. be defined as follows:

123
730 Neural Comput & Applic (2016) 27:727–737

~ BÞ
DEucl ðA; ~ • For cost criteria (smaller the better), dijN ¼ ðdj  dij Þ=
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 )ffi ðdj  djþ Þ.
u n (  
uX TA~ ðxi Þ  TB~ ðxi Þ 2 þ IA~ ðxi Þ  IB~ ðxi Þ 2
¼t  2
i¼1 þ FA~ ðxi Þ  FB~ðxi Þ
ð4Þ 3.2 Step 2. Calculation of weighted normalized
decision matrix
and the normalized Euclidean distance between two
SVNSs A~ and B~ can be defined as follows: In the weighted normalized decision matrix, the modified
~ BÞ
~ ratings are calculated as the following way:
DNEucl ðA;
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
( 2  2 ) vij ¼ wj dijN for i ¼ 1; 2; . . .; m and j ¼ 1; 2; . . .; n:
u
u1 X n T ~ ðx i Þ  T ~ ðx i Þ þ I ~ ðx i Þ  I ~ ðx i Þ
¼t A B

A
2
B ð6Þ
3n i¼1 þ FA~ ðxi Þ  FB~ ðxi Þ
where wj is the weight of the jth criteria such that wj 0 for
ð5Þ P
j ¼ 1; 2; . . .; n and nj¼1 wj ¼ 1.
Definition 8 (Deneutrosophication of SVNS) Deneutro-
sophication of SVNS N~ can be defined as a process of 3.3 Step 3. Determination of the positive
mapping N~ into a single crisp output w 2 X i.e., f : N~ ! and the negative ideal solutions
w for x 2 X. If N~ is discrete set then the vector of tetrads
D E The positive ideal solution (PIS) and the negative ideal
N~ ¼ fðxj T ~ ðxÞ; I ~ ðxÞ; F ~ ðxÞ Þjx 2 Xg is reduced to a
N N N solution (NIS) are derived as follows:
single scalar quantity w 2 X by deneutrosophication. The  
PIS ¼ Aþ ¼ vþ þ þ
1 ; v2 ; . . .vn ;
obtained scalar quantity w 2 X best represents the aggre- ( ! ! )
gate distribution of three membership degrees of neutro- ¼ max vij jj 2 J1 ; min vij jj 2 J2 jj ¼ 1; 2; . . .; n
D E j j
sophic element TN~ ðxÞ; IN~ ðxÞ; FN~ ðxÞ .
ð7Þ
and
3 TOPSIS  
NIS ¼ A ¼ v  
1 ; v2 ; . . .vn ;
( ! ! )
TOPSIS method is used to determine the best alternative ¼ min vij jj 2 J1 ; max vij jj 2 J2 jj ¼ 1; 2; . . .; n
from the concepts of the compromise solution. The best j j
compromise solution should have the shortest Euclidean ð8Þ
distance from the ideal solution and the farthest Euclidean
where J1 and J2 are the benefit and cost-type criteria,
distance from the negative ideal solution. The procedures
respectively.
of TOPSIS can be described as follows. Let A ¼
fA1 ; A2 ; . . .Am g be the set of alternatives, C ¼
 
fC1 ; C2 ; . . .Cn g be the set of criteria and D ¼ dij , 3.4 Step 4. Calculation of the separation measures
i ¼ 1; 2; . . .; m, j ¼ 1; 2; . . .; n, be the performance ratings for each alternative from the PIS and the NIS
with the criteria weight vector W ¼ fwj jj ¼ 1; 2; . . .; ng.
TOPSIS method is presented with these following steps. The separation values for the PIS can be measured by using
the n-dimensional Euclidean distance, which is given as:
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3.1 Step 1. Normalization the decision matrix uX
u n
2
Dþ
i ¼ t v ij  v þ
j i ¼ 1; 2; . . .m: ð9Þ
The normalized value dijN is calculated as follows: j¼1

• For benefit criteria (larger the better), dijN ¼ ðdij  dj Þ= Similarly, separation values for the NIS is
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðdjþ  dj Þ, where djþ ¼ max i ðdij Þ and dj ¼ min i ðdij Þ uX
u n
2
or setting djþ is the aspired or desired level and dj is the Di ¼ t

vij  v j i ¼ 1; 2; . . .m: ð10Þ
j¼1
worst level.

123
Neural Comput & Applic (2016) 27:727–737 731

3.5 Step 5. Calculation of the relative closeness problem. The neutrosophic values associated with the al-
coefficient to the positive ideal solution ternatives for MADM problems can be represented in the
following decision matrix:
The relative closeness coefficient for the alternative Ai with D E  
respect to Aþ is DN~ ¼ dijs ¼ Tij ; Iij ; Fij m n ð13Þ
m n
D
Ci ¼ þ i  for i ¼ 1; 2; . . .m: ð11Þ 0
C1 C2 ... Cn
1
Di þ Di A1 hT11 ; I11 ; F11 i hT12 ; I12 ; F12 i ... hT1n ; I1n ; F1n i
¼ A2 B hT ; I ; F i hT22 ; I22 ; F22 i ... hT2n ; I2n ; F2n i C
B 21 21 21 C
B C
... @ ... ... ... ... A
3.6 Step 6. Ranking the alternatives
Am hTm1 ; Im1 ; Fm1 i hTm2 ; Im2 ; Fm2 i ... hTmn ; Imn ; Fmn i

According to relative closeness coefficient to the ideal alter- ð14Þ

 
native, larger value of Ci indicates the better alternative Ai . In the matrix DN~ ¼ Tij ; Iij ; Fij m n , Tij ,Iij and Fij denote
the degree of truth-membership value, indeterminacy-
membership value and falsity-membership value of alter-
4 TOPSIS method for multi-attribute decision- native Ai with respect to attribute Cj satisfying the fol-
making with single-valued neutrosophic lowing properties under the single-valued neutrosophic
information environment:

Consider a multi-attribute decision-making problem with m 1. 0  Tij  1; 0  Iij  1; 0  Fij  1.

alternatives and n attributes. Let A ¼ fA1 ; A2 ; :::; Am g be a 2. 0  Tij þ Iij þ Fij  3 for i ¼ 1; 2; . . .; n and
discrete set of alternatives, and C ¼ fC1 ; C2 ; . . .; Cn g be j ¼ 1; 2; . . .; m.
the set of attributes. The rating provided by the decision The ratings of each alternative over the attributes are best
maker describes the performance of alternative Ai against illustrated by the neutrosophic cube [24] proposed by
attribute Cj . Let us also assume that W ¼ fw1 ; w2 . . .; wn g Dezert in 2002. The vertices of neutrosophic cube
be the weight vector assigned for the attributes are ð0; 0; 0Þ; ð1; 0; 0Þ; ð1; 0; 1Þ; ð0; 0; 1Þ; ð0; 1; 0Þ; ð1; 1; 0Þ;
C1 ; C2 ; . . .; Cn by the decision makers. The values associ- ð1; 1; 1Þ and ð0; 1; 1Þ. The area of ratings in neutrosophic
ated with the alternatives for MADM problems can be cube is classified in three categories, namely 1. highly
presented in the following decision matrix acceptable neutrosophic ratings, 2. tolerable neutrosophic
C1 C2 ... Cn rating, and 3. unacceptable neutrosophic ratings.
0 1
A1 d11 d12 ... d1n Definition 9 (Highly acceptable neutrosophic ratings)
  Bd
D ¼ dij m n ¼ A2 B 21 d22 ... d2n C
C The subcube (K) of a neutrosophic cube (D) (i.e., K  D)
B C represents the area of highly acceptable neutrosophic rat-
... @ ... ... ... ... A
ings (U)for decision-making. Vertices of K are defined
Am d1m d2m ... dmn
with these eight points ð0:5; 0; 0Þ; ð1; 0; 0Þ; ð1; 0; 0:5Þ;
ð12Þ ð0:5; 0; 0:5Þ; ð0:5; 0; 0:5Þ; ð1; 0; 0:5Þ; ð1; 0:5; 0:5Þ and ð0:5;
4.1 Step 1. Determination of the most important 0:5; 0:5Þ. U includes all the ratings of alternative consid-
attribute ered with the above average truth-membership degree,
below average indeterminacy-membership degree, and
Generally, there are many criteria or attributes in decision- below average falsity-membership degree for multi-at-
making problems, where some of them are important and tribute decision-making. Therefore, U has a great contri-
others may not be so important. So it is crucial to select the bution in decision-making process and can be defined as
proper criteria or attributes for decision-making situation.  
U ¼ Tij ; Iij ; Fij where 0:5\Tij \1; 0\Iij \0:5 and
The most important attributes may be chosen with the help
0\Fij \0:5:
of expert opinions or by some other method that are
technically sound.
for i ¼ 1; 2; . . .; m and j ¼ 1; 2; . . .; n.
4.2 Step 2. Construction of the decision matrix Definition 10 (Unacceptable neutrosophic ratings) The
with SVNSs area C of unacceptable neutrosophic ratings V is defined by
the ratings which are characterized by 0 % membership
It is assumed that the rating of each alternative with respect degree, 100 % indeterminacy degree and 100 % falsity-
to each attribute is expressed as SVNS for MADM membership degree. Thus, the set of unacceptable ratings V

123
732 Neural Comput & Applic (2016) 27:727–737

can be considered as the set of all ratings whose truth- Eq. (15) the weight of the kth decision maker can be
membership value is zero. written as:
  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
V ¼ Tij ; Iij ; Fij where 1  fð1  Tk ðxÞÞ2 þ ðIk ðxÞÞ2 þ ðFk ðxÞÞ2 g=3
Tij ¼ 0; 0\Iij  1 and 0\Fij  1: wk ¼  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pp
k¼1 1  fð1  Tk ðxÞÞ2 þ ðIk ðxÞÞ2 þ ðFk ðxÞÞ2 g=3
for i ¼ 1; 2; . . .; m and j ¼ 1; 2; . . .; n.
Consideration of V should be avoided in decision- ð16Þ
making process. Pp
and k¼1 wk ¼ 1
Definition 11 (Tolerable neutrosophic ratings) Exclud-
ing the area of highly acceptable ratings and unacceptable 4.4 Step 4. Construction of the aggregated single-
ratings from a neutrosophic cube, tolerable neutrosophic valued neutrosophic decision matrix based
rating area H ð¼ D \ :K \ :CÞ can be determined. The on decision makers’ assessments
tolerable neutrosophic rating ðZÞ considered with below
ðkÞ
average truth-membership degree, above average indeter- Let DðkÞ ¼ ðdij Þm n be the single-valued neutrosophic de-
minacy degree and above average falsity-membership de- cision matrix of the k  th decision maker and W ¼
gree are taken in decision-making process. Z can be  T
defined by the following expression w1 ; w2 . . .; wp be the weight vector of decision maker such
  that each wk 2 ½0; 1. In the group decision-making process,
Z ¼ Tij ; Iij ; Fij all the individual assessments need to be fused into a group
where 0\Tij \0:5; 0:5\Iij \1 and 0:5\Fij \1: opinion to make an aggregated neutrosophic decision matrix.
for i ¼ 1; 2; . . .; m and j ¼ 1; 2; . . .; n. This aggregated matrix can be obtained by using single-
D valued neutrosophic weighted averaging (SVNWA) aggre-
Definition 12 Fuzzification of SVNS N~ ¼ fðxj TN~ ðxÞ; gation operator proposed by Ye [17] for SVNSs as follows:
IN~ ðxÞ; FN~ ðxÞiÞjx 2 Xg can be defined as a process of D ¼ ðdij Þm n where,

ð1Þ ð2Þ ðpÞ

mapping N~ into fuzzy set F~ ¼ fxjlF~ðxÞjx 2 Xg i.e., f : dij ¼ SVNSWAW dij ; dij ; . . .; dij
N~ ! F~ for x 2 X. The representative fuzzy membership ð1Þ ð2Þ ðpÞ
¼ w1 dij
w2 dij
  
wðpÞ dij
degree lF~ðxÞ 2 ½0; 11 of the vector tetrads *
p
+
Y
Y p
Y p

fðxjhTN~ ðxÞ; IN~ ðxÞ; FN~ ðxÞiÞjx 2 Xg is defined from the ðpÞ wk ðpÞ wk ðpÞ wk
¼ 1 1  Tij ; Iij ; Fij
concept of neutrosophic cube. It can be obtained by de- k¼1 k¼1 k¼1
termining the root mean square of 1  TN~ ðxÞ, IN~ ðxÞ and ð17Þ
FN~ ðxÞ for all x 2 X. Therefore, the equivalent fuzzy
Therefore, the aggregated neutrosophic decision matrix is
membership degree is as:
defined as follows:
lF~ðxÞ
   
( qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi D ¼ dij m n ¼ Tij ; Iij ; Fij m n ð18Þ
1 fð1  TN~ ðxÞÞ2 þ IN~ ðxÞ2 þ FN~ ðxÞ2 g=3; for 8x 2 U [ Z
¼ C1 C2 ... Cn
0; for 8x 2 V 0 1
A1 hT11 ; I11 ; F11 i hT12 ; I12 ; F12 i ... hT1n ; I1n ; F1n i
ð15Þ ¼ A2 B hT21 ; I21 ; F21 i hT22 ; I22 ; F22 i ... hT2n ; I2n ; F2n i C
B C
B C
... @ ... ... ... ... A
4.3 Step 3. Determination of the weights of decision
Am hTm1 ; Im1 ; Fm1 i hTm2 ; Im2 ; Fm2 i ... hTmn ; Imn ; Fmn i
makers
ð19Þ
 
Let us assume that the group of p decision makers having where, dij ¼ Tij ; Iij ; Fij is the aggregated element of
their own decision weights. Thus, the importance of the neutrosophic decision matrix D for i ¼ 1; 2; . . .m and
decision makers in a committee may not be equal to each j ¼ 1; 2; . . .n.
other. Let us assume that the importance of each decision
maker is considered with linguistic variables and expressed 4.5 Step 5. Determination of the attribute weights
it by neutrosophic numbers.
Let Ek ¼ hTk ; Ik ; Fk i be a neutrosophic number defined In the decision-making process, decision makers may feel
for the rating of kth decision maker. Then, according to that all attributes are not equally important. Thus, every

123
Neural Comput & Applic (2016) 27:727–737 733

decision maker may have their very own opinion regarding In practical, two types of attributes namely, benefit-type
attribute weights. To obtain the grouped opinion of the attribute and cost-type attribute exist in multi-attribute
chosen attribute, all the decision makers’ opinions for the decision-making problem.
importance of each attribute need to be aggregated. Let wkj ¼ Definition 13 Let J1 and J2 be the benefit-type attribute
ð1Þ ð2Þ ðpÞ
ðwj ; wj . . .; wj Þ be the neutrosophic number(NN) as- and cost-type attribute, respectively. Qþ N~
is the relative
signed to the attribute Cj by the kth decision maker. Then the neutrosophic positive ideal solution (RNPIS) and Q N~
is the
combined weight W ¼ fw1 ; w2 . . .; wn g of the attribute can relative neutrosophic negative ideal solution (RNNIS).
be determined by using SVNWA aggregation operator [17]. Then Qþ N~
can be defined as follows:

wþ wþ 
ð1Þ ð2Þ ðpÞ
wj ¼ SVNWAW wj ; wj ; . . .; wj QN~ ¼ d1 ; d2 ; . . .; dnw þ
þ
ð24Þ
ð1Þ ð2Þ ðpÞ D E
¼ w1 wj
w2 wj
  
wðpÞ wj where, djw þ ¼ Tjw þ ; Ijw þ ; Fjw þ for j ¼ 1; 2; . . .; n.
* +
p
Y
Y p
Y p

ðpÞ wk ðpÞ wk ðpÞ wk ( ! !)

¼ 1 1  Tj ; Ij ; Fj
w w
k¼1 k¼1 k¼1 Tjw þ ¼ maxfTij j gjj 2 J1 ; minfTij j gjj 2 J2
i i
ð20Þ
ð25Þ
W ¼ fw1 ; w2 ; . . .; wn g ð21Þ ( ! !)
  w w
where, wj ¼ Tj ; Ij ; Fj for j ¼ 1; 2; . . .n. Ijw þ ¼ minfIij j gjj 2 J1 ; maxfIij j gjj 2 J2 ð26Þ
i i
4.6 Step 6. Aggregation of the weighted ( ! !)
w w
neutrosophic decision matrix Fjw þ ¼ minfFij j gjj 2 J1 ; maxfFij j gjj 2 J2
i i
In this section, the obtained weights of attribute and ag- ð27Þ
gregated neutrosophic decision matrix need to be further
fused to make the aggregated weighted neutrosophic de- and QN~
can be defined by
cision matrix. 
QN~
¼ d1w  ; d2w  ; . . .; dnw  ð28Þ
The aggregated weighted neutrosophic decision matrix
D E
can be obtained by using the multiplication formula (3) of where, djw  ¼ Tjw  ; Ijw  ; Fjw  for j ¼ 1; 2; . . .; n.
two neutrosophic sets as:
D E D E ( ! !)
w w w w
D  W ¼ Dw ¼ dij j ¼ Tij j ; Iij j ; Fij j ð22Þ Tjw  ¼
w
minfTij j gjj 2 J1 ;
w
maxfTij j gjj 2 J2
m n m n
i i
C1 C2 ... Cn
0  w1 w1 w1   w2 w2 w2
  wn wn wn 1
 ð29Þ
A1 T11 ; I11 ; F11 T12 ; I12 ; F12 ... T1n ; I1n ; F1n
B T w1 ; I w1 ; F w1   w2 w2 w2   wn wn wn  C ( ! !)
¼ A2 B 21 21 21 T22 ; I22 ; F22 ... T2n ; I2n ; F2n C
B C w w
... @ ... ... ... ... A Ijw  ¼ maxfIij j gjj 2 J1 ; minfIij j gjj 2 J2
 w1 w1 w1   w2 w2 w2
  w w w i i
Am Tm1 ; Im1 ; Fm1 Tm2 ; Im2 ; Fm2 ... Tmnn ; Imnn ; Fmnn
ð23Þ ð30Þ
D E ( ! !)
w w w w
Here, dij j ¼ Tij j ; Iij j ; Fij j is an element of the aggregated w w
Fjw  ¼ maxfFij j gjj 2 J1 ; minfFij j gjj 2 J2
weighted neutrosophic decision matrix Dw for i ¼ i i
1; 2; . . .; m and j ¼ 1; 2; . . .; n.
ð31Þ
4.7 Step 7. Determination of the relative positive 4.8 Step 8. Determination of the distance measure
ideal solution (RPIS) and the relative negative of each alternative from the RNPIS
ideal solution (RNIS) for SVNSs and the RNNIS for SVNSs
D E  
Let DN~ ¼ dijw ¼ Tij ; Iij ; Fij m n be a SVNS-based Similar to Eq. (5), the normalized Euclidean distance
D E
m n w w w
decision matrix, where, Tij , Iij and Fij are the membership measure of each alternative Tij j ; Iij j ; Fij j from the RNPIS
D E
degree, indeterminacy degree and non-membership degree Tjw þ ; Ijw þ ; Fjw þ for i ¼ 1; 2; . . .; m and j ¼ 1; 2; . . .; n can
of evaluation for the attribute Cj with respect to the alter-
be written as follows:
native Ai .

123
734 Neural Comput & Applic (2016) 27:727–737

Diþ
wj wþ 5.1 Step 1. Determination of the weights of decision
Eu dij ; dj
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi makers
u 8
2
2 9
u n < > T wj ðxj Þ  T w þ ðxj Þ þ I wj ðxj Þ  I w þ ðxj Þ = >
u1 X ij j ij j
¼u t3n
The importance of four decision makers in a selection
j¼1 :
> w 2 >
;
þ Fij j ðxj Þ  Fjw þ ðxj Þ committee may not be equal to each other according their
status. Their decision powers are considered as linguistic
ð32Þ terms expressed in Table 1. The importance of each deci-
similarly, the normalized Euclidean distance measure of sion maker expressed by linguistic term with its corre-
D E sponding SVNN is shown in Table 2. The weight of
w w w
each alternative Tij j ; Iij j ; Fij j from the RNNIS
D E decision maker is determined with the help of Eq. (16) as
Tjw  ; Ijw  ; Fjw  can be written as: follows:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

1  ð0:01 þ 0:01 þ 0:01Þ=3
wj w w1 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

Di
Eu dij ; dj 4  0:03=3  0:1025=3  0:6125=3  0:1025=3
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
8
2
2 9ffi
u
u >
< T
wj
ðx Þ  T w
ðx Þ þ I
wj
ðx Þ  I w
ðx Þ >
= ¼ 0:292
u1 X n ij j j j ij j j j
¼u
t3n

j¼1 :
> 2>
þ F j ðxj Þ  F w  ðxj Þ ;
w
ij j Similarly, other three weights of decision w2 ¼ 0:265,
w3 ¼ 0:178 and w4 ¼ 0:265 can be obtained. Thus, the
ð33Þ
weight vector of the four decision maker is:
W ¼ ð0:292; 0:265; 0:178; 0:265Þ ð35Þ
4.9 Step 9. Determination of the relative closeness
coefficient to the neutrosophic ideal solution
for SVNSs 5.2 Step-2. Construction of the aggregated
neutrosophic decision matrix based
The relative closeness coefficient of each alternative Ai on the assessments of decision makers
with respect to the neutrosophic positive ideal solution QþN~
is defined as follows: The linguistic term along with SVNNs is defined in Table 3

to rate each alternative with respect to each attribute. The
Di
w
dij j ; djw  assessment values of each alternative Ai ði ¼ 1; 2; 3; 4Þ with
Eu
Ci ¼
wj


wj

ð34Þ respect to each attribute Cj ðj ¼ 1; 2; 3; 4; 5; 6Þ provided by
w w
Diþ
Eu dij ; dj þ Di Eu dij ; dj four decision makers are listed in Table 4. Then the ag-
gregated neutrosophic decision matrix can be obtained by
where, 0  Ci  1. fusing all the decision makers’ opinion with the help of
aggregation operator [17] as in Table 5.
4.10 Step 10. Ranking the alternatives By using Eq. (17), the aggregated value of the four
decision makers’ assessment values is arbitrarily chosen as
According to the relative closeness coefficient values larger an illustration for the alternative A1 with respect to the
the values of Ci reflects the better alternative Ai for attribute C1 and shown in Eqs. (36), (37), and (38).
i ¼ 1; 2; . . .; m.
T11 ¼1  ð1  0:90Þ0:292 ð1  0:90Þ0:265 ð1  0:80Þ0:178
ð1  0:80Þ0:265 ¼ 1  0:1359 ¼ 0:8641:
5 Numerical example ð36Þ
Table 1 Linguistic terms for rating of attributes and decision makers
Let us suppose that a group of four decision makers
(DM1 ,DM2 , DM3 , DM4 ) intend to select the most suitable Linguistic terms SVNNs
tablet from the four initially chosen tablet ðA1 ; A2 ; A2 ; A4 Þ
Very good/very important (VG/VI) h0:90; 0:10; 0:10i
by considering six attributes namely: Features C1 , Hard-
Good/important (G/I) h0:80; 0:20; 0:15i
ware C2 , Display C3 , Communication C4 , Affordable Price
Fair/medium (F/M) h0:50; 0:40; 0:45i
C5 , Customer care C6 . Based on the proposed approach
Bad/unimportant (B/UI) h0:35; 0:60; 0:70i
discussed in Sect. 4, the considered problem is solved by
Very bad/very unimportant (VB/VUI) h0:10; 0:80; 0:90i
the following steps:

123
Neural Comput & Applic (2016) 27:727–737 735

Table 2 Importance of decision DM-1 DM-2 DM-3 DM-4

makers expressed with SVNNs
LT VI I M I
e
W h0:90; 0:10; 0:10i h0:80; 0:20; 0:15i h0:50; 0:40; 0:45i h0:80; 0:20; 0:15i

Table 3 Linguistic terms for rating the candidates with SVNNs decision maker is rated with linguistic terms shown in
Table 4. Four decision makers’ opinions need to be ag-
Linguistic terms SVNNs
gregated to determine the combined weight of each at-
Extremely good/high (EG/EH) h1:00; 0:00; 0:00i tribute. The fused attribute weight vector is determined by
Very good/high (VG/VH) h0:90; 0:10; 0:05i using Eq. (20) as follows:
Good/high (G/H) h0:80; 0:20; 0:15i
Medium good/high (MG/MH) h0:65; 0:35; 0:30i 2 3
h0:755; 0:222; 0:217i; h0:887; 0:113; 0:107i;
Medium/fair (M/F) h0:50; 0:50; 0:45i 6 7
6 7
Medium bad/medium law (MB/ML) h0:35; 0:65; 0:60i W ¼ 6 h0:765; 0:226; 0:182i; h0:692; 0:277; 0:251i; 7
4 5
Bad/law (B/L) h0:20; 0:75; 0:80i
Very bad/low (VB/VL) h0:10; 0:85; 0:90i h0:788; 0:200; 0:180i; h0:700; 0:272; 0:244i
Very very bad/low (VVB/VVL) h0:05; 0:90; 0:95i ð40Þ

I11 ¼ð0:10Þ0:292 ð0:10Þ0:265 ð0:20Þ0:178 ð0:20Þ0:265 ¼ 0:1359:

5.4 Step-4. Construction of the aggregated weighted
ð37Þ neutrosophic decision matrix
F11 ¼ ð0:05Þ0:292 ð0:05Þ0:265 ð0:15Þ0:178 ð0:15Þ0:265
After obtaining the combined weights of attribute and
¼ 0:0813: ð38Þ
the ratings of alternative, the aggregated weighted neu-
trosophic decision matrix shown in Table 6 can be
5.3 Step-3. Determine the weights of attribute formed. For example, the element of aggregated
weighted decision matrix for the alternative A1 with re-
The linguistic terms shown in Table 1 are used to evaluate spect to attribute C1 is determined by the following
each attribute. The importance of each attribute for every Eq. (41).

Table 4 Assessments of Alternatives ðAi Þ Decision makers C1 C2 C3 C4 C5 C6

alternatives and attribute
weights given by four decision A1 DM-1 VG G G G G VG
makers
DM-2 VG VG G G G VG
DM-3 G VG G G VG G
DM-4 G G G G G G
A2 DM-1 M G M G G M
DM-2 G MG G G MG G
DM-3 G M G G M M
DM-4 M G M G M M
A3 DM-1 VG VG G G VG VG
DM-2 G VG VG G G VG
DM-3 VG G G MG G MG
DM-4 VG VG G G MG G
A4 DM-1 M VG G G VG M
DM-2 M M G G M G
DM-3 G G G G M VG
DM-4 G M M G G VG
Weights DM-1 VI VI I M I I
DM-2 I VI I I M M
DM-3 M I M M I M
DM-4 M VI M I VI I

123
736 Neural Comput & Applic (2016) 27:727–737

Table 5 Aggregated C1 C2 C3
0 1
neutrosophic decision matrix A1 h0:864; 0:136; 0:081i h0:853; 0:147; 0:092i h0:800; 0:200; 0:150i
A2 B h0:667; 0:333; 0:277i h0:727; 0:273; 0:219i h0:667; 0:333; 0:277i C
B C
A3 @ h0:880; 0:120; 0:067i h0:887; 0:113; 0:064i h0:834; 0:166; 0:112i A
A4 h0:667; 0:333; 0:277i h0:735; 0:265; 0:195i h0:768; 0:232; 0:180i
0 C4 C5 C6 1 (39)
A1 h0:704; 0:296; 0:241i h0:823; 0:177; 0:123i h0:864; 0:136; 0:081i
A2 B h0:744; 0:256; 0:204i h0:652; 0:348; 0:293i h0:608; 0:392; 0:336i C
B C
A3 @ h0:779; 0:221; 0:170i h0:811; 0:189; 0:109i h0:850; 0:150; 0:092i A
A4 h0:727; 0:273; 0:221i h0:791; 0:209; 0:148i h0:808; 0:192; 0:127i

Table 6 Aggregated weighted C1 C2 C3

0 1
neutrosophic decision matrix A1 h0:652; 0:328; 0:280i h0:757; 0:243; 0:289i h0:612; 0:381; 0:305i
A2 B h0:504; 0:481; 0:434i h0:645; 0:355; 0:388i h0:510; 0:484; 0:409i C
B C
A3 @ h0:664; 0:315; 0:269i h0:787; 0:213; 0:267i h0:638; 0:354; 0:274i A
A4 h0:504; 0:481; 0:434i h0:652; 0:348; 0:370i h0:588; 0:406; 0:329i
0 C4 C5 C6 1 (42)
A1 h0:487; 0:491; 0:432i h0:649; 0:342; 0:281i h0:605; 0:371; 0:305i
A2 B h0:515; 0:402; 0:404i h0:514; 0:478; 0:420i h0:426; 0:557; 0:498i C
B C
A3 @ h0:539; 0:437; 0:378i h0:639; 0:351; 0:269i h0:595; 0:381; 0:314i A
A4 h0:503; 0:474; 0:417i h0:623; 0:367; 0:301i h0:566; 0:412; 0:340i

* 0:864 0:755; 0:136 þ 0:222  0:136 0:222; 0:081 + 2 3

h0:504; 0:481; 0:434i; h0:645; 0:355; 0:388i;
w w
hT11 w
; I11 ; F11 i¼ 6 7
Q
N~
¼ 4 h0:510; 0:484; 0:409i; h0:487; 0:491; 0:432i; 5
þ0:217  0:081 0:217
¼h0:65232; 0:3278; 0:2804i h0:514; 0:478; 0:420i; h0:426; 0:557; 0:498i
ð41Þ ð44Þ
 
where, d1w  ¼ T1w  ; I1w  ; F1 w
is calculated as
5.5 Step-5. Determination of the neutrosophic T1w  ¼ minf0:652; 0:504; 0:664; 0:504g ¼ 0:504
relative positive ideal solution
I1w  ¼ maxf0:328; 0:481; 0:315; 0:481g ¼ 0:481
and the neutrosophic relative negative ideal
solution F1w  ¼ maxf0:280; 0:434; 0:269; 0:434g ¼ 0:434

and other components are similarly calculated.

The NRPIS can be calculated from the aggregated
weighted decision matrix on the basis of attribute types, 5.6 Step-6. Determination of the distance measure
i.e., benefit type or cost type by using Eq. (24) as of each alternative from the RNPIS
2 3
h0:664; 0:315; 0:269i; h0:787; 0:213; 0:267i; and the RNNIS and relative closeness coefficient
6 7
QþN~
¼ 4 h0:638; 0:354; 0:274i; h0:539; 0:437; 0:378i; 5
Normalized Euclidean distance measures defined in
h0:649; 0:342; 0:269i; h0:605; 0:371; 0:305i
Eqs. (32) and (33) are used to determine the distances of
ð43Þ
 
where, d1w þ ¼ T1w þ ; I1w þ ; F1 wþ
is calculated as Table 7 Distance measure and relative closeness coefficient of each
alternative
T1w þ ¼ maxf0:652; 0:504; 0:664; 0:504g ¼ 0:664
Alternatives ðAi Þ Diþ
Eucl Di
Eucl Ci
I1w þ ¼ minf0:328; 0:481; 0:315; 0:481g ¼ 0:315
A1 0.0283 0.1281 0.8190
F1w þ ¼ minf0:280; 0:434; 0:269; 0:434g ¼ 0:269
A2 0.3472 0.0490 0.1158
and others. Similarly, the NRNIS can be calculated from A3 0.0224 0.1382 0.8605
aggregated weighted decision matrix on the basis of at- A4 0.0900 0.0831 0.4801
tribute types, i.e., benefit type or cost type by using The largest relative closeness value expressed in bold indicates that
Eq. (28) as A3 is the mostdesirable alternative

123
Neural Comput & Applic (2016) 27:727–737 737

each alternative from the RNPIS and the RNNIS. With 4. Roy B (1991) The outranking approach and the foundations of
these distances, relative closeness coefficient is calculated ELECTRE methods. Theory Decis 31:49–73
5. Zadeh LA (1965) Fuzzy sets. Inf Control 8:338–353
by using Eq. (34). These results are listed in Table 7. 6. Chen CT (2000) Extensions of the TOPSIS for group decision
making under fuzzy environment. Fuzzy Sets Syst 114:1–9
5.7 Step-7. Ranking the alternatives 7. Atanassov KT (1986) Intuitionistic fuzzy sets. Fuzzy Sets Syst
20:87–96
8. Boran FE, Genc S, Kurt M, Akay D (2009) A multi-criteria in-
According to the values of relative closeness coefficient of tuitionistic fuzzy group decision making for supplier selection
each alternative shown in Table 7, the ranking order of four with TOPSIS method. Expert Syst Appl 36(8):11363–11368
alternatives is 9. Pramanik S, Mukhopadhyaya D (2011) Grey relational analysis
based intuitionistic fuzzy multi-criteria group decision-making
A3  A1  A4  A2 : approach for teacher selection in higher education. Int J Comput
Appl 34(10):21–29
Thus, A3 is the best alternative tablet. 10. Smarandache F (1999) A unifying field in logics. Neutrosophy:
neutrosophic probability, set and logic. American Research Press,
Rehoboth
6 Conclusions 11. Wang H, Smarandache F, Zhang YQ, Sunderraman R (2010)
Single valued neutrosophic sets. Multispace Multistruct
4:410–413
This paper is devoted to present a new TOPSIS-based 12. Ye J (2013) Multicriteria decision-making method using the
approach for MAGDM under simplified neutrosophic en- correlation coefficient under single-valued neutrosophic envi-
vironment. In the evaluation process, the ratings of each ronment. Int J Gen Syst 42(4):386–394
13. Ye J (2013) Single valued neutrosophic cross entropy for multi-
alternative with respect to each attribute are given as lin- criteria decision making problems. Appl Math Model
guistic variables characterized by single-valued neutro- 38:1170–1175
sophic numbers. Neutrosophic aggregation operator is used 14. Biswas P, Pramanik S, Giri BC (2014) Entropy based grey re-
to aggregate all the opinions of decision makers. Neutro- lational analysis method for multi-attribute decision-making un-
der single valued neutrosophic assessments. Neutrosophic Sets
sophic positive ideal and neutrosophic negative ideal so- Syst 2:102–110
lution are defined from aggregated weighted decision 15. Biswas P, Pramanik S, Giri BC (2014) A new methodology for
matrix. Euclidean distance measure is used to determine neutrosophic multi-attribute decision making with unknown
the distances of each alternative from positive as well as weight information. Neutrosophic Sets Syst 3:42–52
16. Zhang HY, Wang JQ, Chen XH (2014) Interval neutrosophic sets
negative ideal solutions for relative closeness coefficient of and their application in multi-criteria decision making problems.
each alternative. However, the author hopes that the con- Sci World J. doi:10.1155/2014/645953
cept presented in this paper may open up new avenue of 17. Ye J (2014) A multi-criteria decision-making method using ag-
research in competitive neutrosophic decision-making gregation operators for simplified neutrosophic sets. J Intell
Fuzzy Syst 26:2459–2466
arena. TOPSIS method with neutrosophic set information 18. Chi P, Liu P (2013) An extended TOPSIS method for the multi-
has enormous chance of success for multi-attribute deci- attribute decision making problems on interval neutrosophic set.
sion-making problems. In future, the proposed approach Neutrosophic Sets Syst 1:63–70
can be used for dealing with decision-making problems 19. Rivieccio U (2008) Neutrosophic logics: prospects and problems.
Fuzzy Sets Syst 159:1860–1868
such as personal selection in academia, project evaluation, 20. Ghaderi SF, Azadeh A, Nokhandan BP, Fathi E (2012) Behav-
supplier selection, manufacturing systems, and many other ioral simulation and optimization of generation companies in
areas of management systems. electrical markets by fuzzy cognitive map. Expert Syst Appl
39:4635–4646
21. Smarandache F (2005) Neutrosophic set: a generalization of the
intuitionistic fuzzy set. Int J Pure Appl Math 24:287–297
22. Liu P, Wang Y (2014) Multiple attribute decision making method
References based on single-valued neutrosophic normalized weighted Bon-
ferroni mean. Neural Comput Appl 25:2001–2010
1. Hwang CL, Yoon K (1981) Multiple attribute decision making: 23. Majumdar P, Samanta SK (2013) On similarity and entropy of
methods and applications. Springer, New York neutrosophic sets. J Intell Fuzzy Syst. doi:10.3233/IFS-130810
2. Brans JP, Vinvke P, Mareschal B (1986) How to select and how 24. Dezert J (2002) Open questions in neutrosophic inferences. Mult
to rank projects: the PROMETHEE method. Eur J Oper Res Valued Log Intern J 8:439–472
24:228–238
3. Opricovic S, Tzeng GH (2004) Compromise solution by MCDM
methods: a comparative analysis of VIKOR and TOPSIS. Eur J
Oper Res 156:445–455

123