Bipolar Neutrosophic Dombi Aggregation Operators and Their Application To Multi-Attribute Decision-Making Problems

You might also like

You are on page 1of 35

Bipolar Neutrosophic Dombi Aggregation

Operators and Their Application to


Multi-attribute Decision-Making Problems

a. Department of Mathematics, University of the Punjab, New Campus,


Lahore, Pakistan.
E-mail:gulfamhanifg@gmail.com

Abstract
In this paper, Dombi t-norm(TC) and Dombi t-conorm(TCN) are used to gener-
ate more complex, flexible and feasible operation rules by managing a parameter in
bipolar neutrosophic fuzzy(BNF) environment. We introduce bipolar neutrosophic
Dombi weighted geometric aggregation(BNDWGA) operator, bipolar neutrosophic
Dombi ordered weighted geometric aggregation(BNDOWGA) operator, bipolar neu-
trosophic Dombi prioritized weighted geometric aggregation(BNDPWGA) opera-
tor, bipolar neutrosophic Dombi prioritized ordered weighted geometric aggregation
(BNDPOWGA) operator along with proofs to aggregate suitable preferences of de-
cision maker. We investigate different properties for all operators. Multi-attribute
decision making(MADM) methods are developed based on proposed aggregation
operators under bipolar neutrosophic fuzzy(BNF) environment. A numerical exam-
ple of selection of cultivating crop is used to explain the proposed methods under
bipolar neutrosophic fuzzy(BNF) environment. The effect of parameters are also
discussed. Finally, the proposed methods are compared with the existing methods
to verify their suitable preferences.

Key-Words: Bipolar neutrosophic set, Aggregation operators, Multi-attribute and


Decision-Making.

1 Introduction
The representation of decision-making information has varied which is no longer limited
to real numbers due to the increasing uncertainty and complexity in decision-making en-
vironment. Fuzzy numbers have been used to characterize the fuzzy information instead

1
of real numbers. Zadeh [1] has introduced the concept of fuzzy set(FS). Nevertheless,
FSs cannot solve complex problems because they all have only one membership degree.
Atonasso [2] has introduced the concept of intuitionistic fuzzy set(IFS) for a highly flexi-
ble description of uncertain information. IFS includes both membership degree and non-
membership degree. Thereafter, Atanassov and Gargov [3] have introduced the concept of
interval-valued intuitionistic fuzzy set(IVIFS) which is a generalization of IFS. In practi-
cal decision-making problems, IFS and IVIFS can not represent inconsistent information.
Smarandache [4] introduced neutrosophic set which includes truth membership degree,
indeterminacy degree, and falsity membership degree to depict incomplete, inconsistent
and uncertain information. Wang et al. [5, 6] introduced the interval neutrosophic set
(INS) and the single-valued neutrosophic set (SVNS) to apply NSs in practical problems.
Positive and negative effects can be created in human minds as we make a decision.
Positive information explains what is possible, satisfactory, permitted, described, or ac-
ceptable. On contrary, negative information explains what is impossible, rejected, or
forbidden [7]. Satisfactory or acceptable conception refers to positive preferences, while
unsatisfactory or unacceptable conception refers to negative preferences. Positive prefer-
ences corresponds to wishes, while negative preferences correspond to constraints [8]. For
example, when a decision maker(DM) evaluates an object, he may explain what he con-
siders (more or less) satisfactory or acceptable; on contrary, he may also explain what he
considers unsatisfactory or unacceptable because of different constraints [9]. To describe
the aforesaid information, Zhang [10] introduced the bipolar fuzzy set which consists of
the following two parts: positive membership degree and negative membership degree.
Given the advantages of BFS and NS, Deli et al. [11] introduced the bipolar neutrosophic
set(BNS) that can describe fuzzy, bipolar, uncertain and inconsistent information. Dey et
al.[12] proposed the bipolar neutrosophic TOPSIS(BN-TOPSIS) method to solve MADM
problems under bipolar neutrosophic fuzzy environment. Zhang et al. [9] proposed the
methods based on the Frank choquet Bonferroni Mean Operators to solve MADM prob-
lems under bipolar neutrosophic fuzzy environment.
Multi-attribute decision-making (MADM) refers to make decisions when there are multi-
ple but a finite list of alternatives and attributes. Dombi [13] introduced the new triangular
norms which are Dombi TN and Dombi TCN. Dombi TN and Dombi TCN show the good
flexibility with operational parameter. Dombi operations till cannot extend to aggregate
bipolar neutrosophic fuzzy environment. In this paper, we developed some new aggre-
gation operators based on the combination of bipolar neutrosophic numbers(BNNs) and
Dombi operations.To propose some bipolar neutrosophic aggregation operators for the
aggregation of bipolar neutrosophic fuzzy information, and to develop MADM methods
based on BNDWGA, BNDOWGA, BNDPWGA, and BNDPOWGA operators to solve
MADM problems with bipolar neutrosophic fuzzy information.
The remainder of this paper is organized as follows. Section 2 reviews several basic def-
initions and concepts which will be used in this paper. Section 3 defines the Dombi
operations of bipolar neutrosophic numbers(BNNs). Section 4 defines the proposed new
BNDWGA and BNDOWGA operators and their properties. Section 5 defines the pro-

2
posed new BNDPWGA and BNDPOWGA operators and their properties. Section 6
proposes comprehensive MADM methods based on the proposed operators. Section 7
provides a numerical example of selection of cultivating crop. Section 8 discusses the ef-
fect of parameter and demonstrates a comparative analysis with other methods. Finally,
section 9 concludes this paper.

2 preliminaries
First we discuss some basic definitions like bipolar neutrosophic set(BNS), Bipolar neu-
trosophic number(BNN), Dombi operations. These definitions will be described in sub-
sections.

2.1 Bipolar Neutrosophic set and Bipolar Neutrosophic Num-


ber

0.55

0.50

0.45

0.40

0.35

0.30

0.25
A1 A2 A3 A4

Alternatives

Figure 1: BNDPWGA OPERATOR.

3
0.55

0.50

0.45

0.40

0.35

0.30

0.25
A1 A2 A3 A4

Alternatives

Figure 2: BNDPOWGA OPERATOR.

4
0.55

0.50

0.45

0.40

0.35

0.30

0.25
A1 A2 A3 A4
Alternatives

Figure 3: BNDWGA OPERATOR.

5
0.55

0.50

0.45

0.40

0.35

0.30

0.25
A1 A2 A3 A4

Alternatives

Figure 4: BNDOWGA OPERATOR.

Definition 2.1. [11] Let X be a fixed set. Then, BNS B can be defined as follows
− − −
B (x), µB (x), νB (x), tB (x), µB (x), νB (x) > |x ∈ X},
B(x) =< t+ + +

− − −
where t+ B (x), µB (x), νB (x) : X −→ [0, 1] and tB (x), µB (x), νB (x) : X −→ [−1, 0]. The pos-
+ +

itive membership degrees t+ + +


B (x), µB (x), νB (x) are the truth membership, indeterminacy
membership degree and falsity membership degree of an element x ∈ X corresponding to
BNS B and the negative membership degrees t− − −
B (x), µB (x), νB (x) denote the truth mem-
bership degree, indeterminacy membership degree and falsity membership degree of an
element x ∈ X to some implicit counter property corresponding to a BNS B.
− − −
In particular, if X has only one element, then B(x) =< t+ + +
B (x), µB (x), νB (x), tB (x), µB (x), νB (x) >
is called a bipolar neutrosophic number(BNN). For convenience, a BNN B =< tB (x), µB (x), νB+ (x),
+ +

t− − − + + + − − −
B (x), µB (x), νB (x) > is also denoted as B =< tB , µB , νB , tB , µB , νB >.

Deli et al [11] defined the algebraic operations as follows:

+ − − − + − − −
Definition 2.2. Let b̃1 =< t+ + + +
1 , µ1 , ν1 , t1 , µ1 , ν1 > and b̃2 =< t2 , µ2 , ν2 , t2 , µ2 , ν2 > be
two BNNs. Then the operations of BNNs are defined as:

6
+ − − − − − − −
(1) b˜1 ⊕ b˜2 =< t+1 + t2 − t1 t2 , µ1 µ2 , ν1 ν2 , −t1 t2 , −(−µ1 − µ2 − µ1 µ2 ), −(−ν1 − ν2 −
+ + + + + + +

ν1− ν2− ) >,


− − − − − − − −
(2) b˜1 ⊗b˜2 =< t+
1 t2 , µ1 +µ2 −µ1 µ2 , ν1 +ν2 −ν1 ν2 , −(−t1 −t2 −t1 t2 ), −µ1 µ2 , −ν1 ν2 >,
+ + + + + + + + +

− k − k − k
(3) k.b˜1 =< 1−(1−t+
1 ) , (µ1 ) , (ν1 ) , (−t1 ) , −(1−(1−(−µ1 ) )), −(1−(1−(−ν1 ) )) >
k + k + k

(k > 0),
k − k − k − k
(4) b˜1 =< (t+ 1 ) , 1 − (1 − µ1 ) , 1 − (1 − ν1 ) , −(1 − (1 − (−t1 ) )), −(−µ1 ) , −(−ν1 ) >
k + k + k

(k > 0).

Deli et al [11] proposed a comparison method in which includes score, accuracy and cer-
tainty function to compare two BNNs.

+ − − −
Definition 2.3. [11] Let b˜1 =< t+ +
1 , µ1 , ν1 , t1 , µ1 , ν1 > be BNN, then the score function
S(b˜1 ), accuracy function A(b˜1 ) and certainty function C(b˜1 ) are defined as:

− − −
1 + 1 − µ1 + 1 − ν1 + 1 + t1 − µ1 − ν1
t+ + +
S(b˜1 ) = (1)
6
˜ − −
A(b1 ) = (t1 − ν1 ) + (t1 − ν1 )
+ +
(2)

C(b˜1 ) = t+
1 − ν1 (3)

The ranking methods of BNNs can be derived based on Eqs. (1) − (3) as follows.
+ − − − + − − −
Definition 2.4. [11] Let b˜1 =< t+ + ˜ + +
1 , µ1 , ν1 , t1 , µ1 , ν1 > and b2 =< t2 , µ2 , ν2 , t2 , µ2 , ν2 >
be two BNNs, therefore

(1) if S(b˜1 ) > S(b˜2 ), then b˜1 > b˜2 ,

(2) if S(b˜1 ) = S(b˜2 ) and A(b˜1 ) > A(b˜2 ) , then b˜1 > b˜2 ,

(3) if S(b˜1 ) = S(b˜2 ), A(b˜1 ) = A(b˜2 ) and C(b˜1 ) > C(b˜2 ), then b˜1 > b˜2 ,

(4) if S(b˜1 ) = S(b˜2 ), A(b˜1 ) = A(b˜2 ) and C(b˜1 ) = C(b˜2 ), then b˜1 ∼ b˜2 .

2.2 Dombi Operations


Dombi product and Dombi sum which are special cases of TN and TCN given in the
following definition.

7
Definition 2.5. [13] Let b1 and b2 be any two real numbers, then Dombi TN and Dombi
TCN are defined in the following expressions:

1
b1 ⊗D b2 = Dom(b1 , b2 ) = 1 (4)
1+ {( 1−b
b1
1 l
) + ( 1−b
b2
)}
2 l l

and
1
b1 ⊕D b2 = Dom∗ (b1 , b2 ) = 1 − 1 (5)
1 + {( 1−b
b1 l
1
)}
b2 l l
) + ( 1−b2

where l ≥ 1 and (b1 , b2 ) ∈ [0, 1] × [0, 1].


Some special cases can be easily proved.

b1 +b2 −2b1 b2
(1) if l −→ 1, then b1 ⊕D b2 −→ 1−b1 b2
and b1 ⊗D b2 −→ b1 b2
b1 +b2 −b1 b2
,

(2) If l −→ ∞, then b1 ⊕D b2 −→ max(b1 , b2 ) and b1 ⊗D b2 −→ min(b1 , b2 ). Dombi sum


and Dombi product are reduced to simple max-operator and simple min-operator,
respectively.

3 Dombi Operations of BNNs


In this section, we defines the Dombi operations of BNNs and discuss their properties.
Dombi operations of BNNs are defined as follows.
+ − − − + − − −
Definition 3.1. Let b̃1 =< t+ + + +
1 , µ1 , ν1 , t1 , µ1 , ν1 > and b̃2 =< t2 , µ2 , ν2 , t2 , µ2 , ν2 > be
two BNNs and l ≥ 1. Then, Dombi product and Dombi sum of two BNNs b̃1 and b̃2 are
denoted by b̃1 ⊗D b̃2 and b̃1 ⊕D b̃2 respectively and defined as follows:



 ⟨1 − 1
, 1
, 1
,
 1+{(
t+ + 1
1 )l +( t2 )l } l
+ 1+{(
1−µ+ + 1
1 )l +( 1−µ2 )l } l 1+{(
1−ν1+ 1−ν + 1
)l +( +2 )l } l
1−t+ + + +
(1) b̃1 ⊕D b̃2 = 1−t1 2 µ1 µ2 ν1 ν2

 −1
 1+{( 1+t−1 )l +( 1+t−2 )l } 1l , 1+{( −µ−1 )l +( −µ−2 )l } 1l − 1, 1+{( −ν1− )l +( −ν2− )l } 1l − 1⟩,
1 1

−t−
1 −t−
2 1+µ−
1 1+µ−
2

1+ν1 −
1+ν2



 ⟨ 1
,1 − 1
,1 − 1
,
 1+{( 1−t+1 )l +( 1−t+2 )l } 1l
+ + µ+ + l
1
1 )l +( µ2 )l } l ν+ +
ν2 1
+) }
1+{( + 1+{( 1 + +( l l
1−µ+
(2) b̃1 ⊗D b̃2 =
t1 t2 1−µ1 2 1−ν1 1−ν2

 −1 −1
 1+{( −t−1 )l +( −t−2 )l } 1l − 1, 1+{( 1+µ−1 )l +( 1+µ−2 )l } 1l , 1+{( 1+ν1− )l +( 1+ν2− )l } 1l ⟩,
1

1+t−
1 1+t−
2 −µ−
1 −µ−
2

−ν1 −
−ν2

8


 ⟨1 − 1
, 1
, 1
,
 1+{k(
t+ 1
1 )l } l 1+{k(
1−µ+ 1
1 )l } l 1+{k(
1−ν1+ 1
)l } l
1−t+ µ+ +
ν1
(3) k.D b̃1 = 1 1
 −1
 1+{k( 1+t−1 )l } 1l , 1+{k( −µ−1 )l } 1l − 1, 1+{k( −ν1− )l } 1l − 1⟩(k > 0),
1 1

−t−
1 1+µ−
1

1+ν1


⟨
 ,1 − ,1 −
1 1 1
,
∧k  1−t+ 1
1+{k( +1 )l } l
µ+ 1
1+{k( 1 + )l } l 1+{k(
+
ν1
+) }
l l
1
t1 1−µ1 1−ν1
(4) b̃1 D =
 −1 −1
 1+{k( −t−1 )l } 1l − 1, 1+{k( 1+µ−1 )l } 1l , 1+{k( 1+ν1− )l } 1l ⟩(k > 0).
1

1+t−
1 −µ−
1

−ν1

+ − − − + − − −
Theorem 3.2. Let b̃1 =< t+ + + +
1 , µ1 , ν1 , t1 , µ1 , ν1 > and b̃2 =< t2 , µ2 , ν2 , t2 , µ2 , ν2 ∧>
k
be two BNNs and let c̃ = b̃1 ⊕D b̃, d˜ = b̃1 ⊗D b̃2 , ẽ = k.D b̃1 (k > 0) and f˜ = b̃1 D
˜ ẽ and f˜ are also BNNs.
(k > 0).Therefore, c̃, d,
As Theorem (3.2) can be easily verified.Therefore, the proof is omitted here.

The properties of Dombi operations of BNNs will be defined as follows.


+ − − − + − − −
Theorem 3.3. Let b̃1 =< t+ + + +
1 , µ1 , ν1 , t1 , µ1 , ν1 > and b̃2 =< t2 , µ2 , ν2 , t2 , µ2 , ν2 > be
two BNNs and k, k1 , k2 > 0.Then,the following properties can be proven easily.

(a) b̃1 ⊕D b̃2 = b̃2 ⊕D b̃1 ,

(b) b̃1 ⊗D b̃2 = b̃2 ⊗D b̃1 ,

(c) k.D (b̃1 ⊕D b̃2 ) = k.D b̃2 ⊕D k.D b̃1 ,


∧k ∧k ∧k
(d) (b̃1 ⊗D b̃2 ) D = b̃1 D
⊗D b̃2 D ,

(e) (k1 + k2 ).D b̃1 = k1 .D b̃1 ⊕D k2 .D b̃1 ,


∧k1 +k2 ∧k1 ∧k 2
(f) b̃1 D = b̃1 D ⊗D b̃1 D .

4 Bipolar Neutrosophic Dombi aggregation opera-


tors
This section proposes the bipolar neutrosophic Dombi weighted geometric aggregation(BNDWGA)
and the bipolar neutrosophic Dombi ordered weighted geometric aggregation(BNDOWGA)
operators and discusses the different properties of these aggregation operators (AOs).

9
4.1 Bipolar neutrosophic Dombi weighted geometric aggrega-
tion operator
+ − − −
Definition 4.1. Let b˜i =< t+ +
i , µi , νi , ti , µi , νi > (i = 1, , 2, 3, ..., n) be a family of
BNNs. A mapping BNDWGA: U n → U is called BNDWGA operator, if it satisfies

n ∧w
i
∧w1 ∧w2 ∧ wn
BNDWGA(b̃1 , b̃2 , ..., b̃n ) = b̃i = b̃1 D ⊗D b̃2 D ⊗D ... ⊗D b̃n D (6)
i=1

where w = (w1 , w2 , ..., wn ) is the weight vector of b̃i , wi ∈ [0, 1] and Σi=1 wi = 1.
T

+ − − −
Theorem 4.2. Let b˜i =< t+ +
i , µi , νi , ti , µi , νi > (i = 1, 2, 3, ..., n) be a family of BNNs
and w = (w1 , w2 , ..., wn )T be the weight vector of bi , wi ∈ [0, 1] and Σi=1 wi = 1. Then,
the aggregated value of BNDWGA operator is still a BNN, which is calculated by the
following formula
∧w1 ∧w2 ∧wn
BN DW GA(b̃1 , b̃2 , ..., b̃n ) = b̃1 D ⊗D b̃2 D ⊗D ... ⊗D b̃n D


 ⟨ 1
,1 − 1
,1 − 1
,
 1+{Σni=1 wi ( 1−t+i )l } 1l
+ µ+ 1 ν+ 1
n
1+{Σi=1 wi ( i
+ ) }l
l 1+{Σi=1 wi ( i + )l } l
n
t 1−µ 1−ν
= i i i
(7)


1
− 1, −1
, −1

 1+{Σn wi ( −ti )l } 1l

1+{Σn wi (
1+µ − 1
i )l } l 1+{Σn wi (
1+ν − 1
i )l } l
i=1 1+t− i=1 −µ− i=1 −ν −
i i i

By the mathematical induction, we can prove theorem (4.2).

Proof: If n = 2 based of the Dombi operations on BNNs in definition (4.1), we can


obtain the following result: ∧w1 ∧w 2
BNDWGA
 ( b̃ 1 , b̃ 2 ) = b̃ 1
D
⊗ D b̃ 1
D


 ⟨ 1
,1 − 1
,1 − 1
,
 1+{w1 ( 1−t++1 )l +w2 ( 1−t++2 )l } 1l µ+ µ+
1+{w1 ( 1 + )l +w2 ( 2 + )l } l
1 ν+ ν+
1+{w1 ( 1 + )l +w2 ( 2 + )l } l
1
t1 t2 1−µ1 1−µ2 1−ν1 1−ν2
=


1
− 1, −1
, −1
⟩,
 1+{w1 ( −t−1 )l +w2 ( −t−2 )l } 1l 1+{w1 (
1+µ− − 1
1 )l +w ( 1+µ2 )l } l 1+{w1 (

1+ν1
)l +w2 (

1+ν2 1
)l } l
− − −µ1 − 2
−µ2− −ν1− −ν2−
 1+t1 1+t2


 ⟨ 1
,1 − 1
,1 − 1
,
 1+{Σ2i=1 wi ( 1−t+i )l } 1l
+ µ+ 1 ν+ 1
1+{Σ2i=1 wi ( i + )l } l 1+{Σ2i=1 wi ( i + )l } l
t 1−µ 1−ν
= i i i


1
− 1, −1
, −1
⟩.
 1+{Σ2 wi ( −ti )l } 1l

1+{Σ2 wi (
1+µ − 1
i )l } l 1+{Σ2 wi (
1+ν − 1
i )l } l
i=1 1+t− i=1 −µ− i=1 −ν −
i i i

If n = m, based on equation (7), then we have the following equation:




 ⟨ 1
,1 − 1
,1 − 1

⊗m ∧ wi
 m 1−t+
1+{Σi=1 wi ( + ) }
i l
1
l m
1+{Σi=1 wi (
µ+
i )}
l
1
l m ν+
1+{Σi=1 wi ( i + )l }
t 1−µ+ 1−ν
BNDWGA(b̃1 , b̃2 , ..., b̃m ) = i=1 b̃i = i i i


1
− 1, −1
, −1
⟩.
 1+{Σm wi ( −ti )l } 1l

1+{Σm wi (
1+µ − 1
i )l } l 1+{Σm wi (
1+ν − 1
i )l } l
i=1 1+t− i=1 −µ− i=1 −ν −
i i i
If n = m + 1, then there is following result:

10
⊗ ∧w ⊗ ∧w
i m+1
BNDWGA(b̃1 , b̃2 , ..., b̃m , b̃m+1 ) = m b̃ b̃m+1
 i=1 i

 ⟨ 1
,1 − 1
,1 − 1
,
 1+{Σm 1−t+ 1 + 1 + 1
w ( i )l } l 1+{Σ m w ( µi )l} l 1+{Σ m w ( νi )l } l
i=1 i t+ i=1 i
1−µ + i=1 i
1−ν +
= i i i


1
− 1, 1
, 1

 1+{Σm wi ( −ti )l } 1l
− 1+µ − 1 1+ν − 1
1+{Σmi=1 wi ( −µ− ) } 1+{Σm i=1 wi ( −ν − ) }
i l l i l l
1+t−

i=1
i i i

 ⟨ 1
,1 − 1
,1 − 1
,
 1−t+
 1+{wm+1 ( + )l } l m+1 1
1+{wm+1 (
µ+ m+1 1
)l } l 1+{wm+1 ( m+1
ν+ 1
)l } l
+ +
⊗D
tm+1 1−µm+1 1−νm+1



1
− 1, 1
, 1


 1+{wm+1 ( −tm+1
− 1
)l } l 1+{wm+1 (
1+µ− m+1 l 1
) } l 1+{wm+1 (
1+νm+1− 1
)l } l
− −µm+1− −
−νm+1

1+tm+1


 ⟨ 1
,1 − 1
,1 − 1
,
 1+{Σm+1i=1 w i (
1−t+
+
i )l } l
1
1+{Σ m+1
i=1 w i (
µ+
i
+ )l} l
1
1+{Σ m+1
i=1 w i(
ν+
i
+
1
)l } l
t 1−µ 1−ν
= i i i


1
− 1, −1
, −1
⟩.
 1+{Σm+1 wi ( −t−i )l } 1l 1+{Σm+1 wi (
1+µ−
i )l } l
1
1+{Σm+1 wi (
1+ν −
i )l } l
1
i=1 1+t− i=1 −µ− i=1 −ν −
i i i

Hence, Theorem (4.2) is true for n = m + 1. Thus, equation (7) holds for all n.

Then, the BNDWGA operator contains the following properties:


+ − − −
(1) Idomopotency: Let all the BNNs be b˜i =< t+ +
i , µi , νi , ti , µi , νi >= b̃(i = 1, 2, 3, ..., n),
where w = (w1 , w2 , ..., wn )T be the weight vector of b̃i , wi ∈ [0, 1] and Σi=1 wi = 1.
Then, BNDWGA(b̃1 , b̃2 , ..., b̃n ) = b̃.

(2) Monotonicity: Let b̃i (i = 1, 2, ..., n) and b̃,i (i = 1, 2, ..., n) be two families of BNNs,
where w = (w1 , w2 , ..., wn )T be the weight vector of b̃i and b̃,i , wi ∈ [0, 1] and Σi=1 wi =
1. For all i ,if b̃i ≥ b̃,i , then BNDWGA(b̃1 , b̃2 , ..., b̃n ) ≥ BNDWGA(b̃,1 , b̃,2 , ..., b̃,n ).
+ − − −
(3) Boundedness: Let b˜i =< t+ +
i , µi , νi , ti , µi , νi > (i = 1, 2, 3, ..., n) be a family of
BNNs, where w = (w1 , w2 , ..., wn ) be the weight vector of bi , wi ∈ [0, 1] and
T

Σni=1 wi = 1. Therefore, we have


BNDWGA(b̃− , b̃− , ..., b̃− ) ≤ BNDWGA(b̃1 , b̃2 , ..., b̃n ) ≤ BNDWGA(b̃+ , b̃+ , ..., b̃+ ),
where b̃− =< t+ , µ+ , νb̃+− , t− , µ− , νb̃−− >
{ b̃− b̃− b̃− b̃−
< min(t+ + + + + + + +
1 , t2 , ...tn , ), max(µ1 , µ2 , ...µn ), max(ν1 , ν2 , ..., νn ),
+
=
max(t− − − − − − − −
1 , t2 , ...tn ), min(µi , µi , ...µi ), min(ν1 , ν2 , ..., νn ) >

and
b̃+ =< t+ , µ+ , νb̃++ , t− , µ− , νb̃−+ >
{ b̃+ b̃+ b̃+ b̃+
< max(t+ + + + + + + +
1 , t2 , ...tn , ), min(µ1 , µ2 , ...µn ), min(ν1 , ν2 , ..., νn ),
+
=
min(t− − − − − − − −
1 , t2 , ...tn ), max(µi , µi , ...µi ), max(ν1 , ν2 , ..., νn ) > .

+ − − −
Proof (1) Since b˜i =< t+ +
i , µi , νi , ti , µi , νi >= b̃(i = 1, , 2, 3, ..., n). Then, by using (7),
we can obtain the following result:

11


 ⟨ 1
,1 − 1
,1 − 1
,
 1+{Σn
i=1 wi (
1−t+ 1
i )l } l 1+{Σn
i=1 wi (
µ+
i
1
)l } l 1+{Σn
i=1 wi (
ν+
i
1
)l } l
t+ 1−µ+ 1−ν +
BNDWGA(b̃1 , b̃2 , ..., b̃n ) = i i i


1
− 1, −1
, −1

 1+{Σn −t− 1 1+µ− 1 1+ν − 1
i=1 wi ( 1+t− ) } i=1 wi ( −µ− ) }
1+{Σn i=1 wi ( −ν − ) }
1+{Σn
i l l i l l i l l
i i i



⟨ 1+{( 1−t+ )l } 1l , 1 −
1 1
µ+ 1 ,1 − 1
ν+
1 ,
1+{( )l } l 1+{( )l } l
= t+ 1−µ+ 1−ν +
=< t+ , µ+ , ν + , t− , µ− , ν − >= b̃.
 −1 −1
 1+{( −t− )l } 1l − 1, 1+{( 1+µ− )l } 1l , 1+{( 1+ν − )l } 1l ⟩
1

1+t− −µ− −ν −

Hence, BNDWGA(b̃1 , b̃2 , ..., b̃n ) = b̃ holds.

(2) based on the equation (7), the property is obvious.

(3) Let b̃− =< t+ b̃−


, µ+
b̃−
, νb̃+− , t−
b̃−
, µ−
b̃−
, νb̃−− > and b̃+ =< t+b̃+
, µ+
b̃+
, νb̃++ , t−
b̃+
, µ−
b̃+
, νb̃−+ >. There
are the following inequalities:
1
1−t+ 1
≤ 1
1−t+ 1
≤ 1
1−t+
,
b̃− )l } l 1+{Σn i )l } l + l 1

1+{Σn
i=1 wi ( i=1 wi ( 1+{Σn ) }l
1+ t+
i i=1 wi ( t+
b̃− b̃+
1− µ+
1
1
≤1− µ+
1
1
≤1− µ+
1
1
,
b̃−
i=1 wi ( 1−µ+ ) }
1+{Σn
b̃+ i l l
i=1 wi ( 1−µ+ ) }
1+{Σn i=1 wi ( 1−µ+ ) }
1+{Σn
l l l l
i
b−
b̃+ ˜

1− ν+
1
1
≤1− 1
ν+ 1
≤1− ν+
1
1
,
b̃+ 1+{Σn i )l } l b̃−
i=1 wi ( 1−ν + ) }
1+{Σn i=1 wi ( i=1 wi ( 1−ν + ) }
1+{Σn
l l l l
1−ν +
i −
b̃+ b˜
1
−t−
−1≤ −t−
1
1
−1≤ 1
−t−
− 1,
i=1 wi ( 1+t− ) }
b˜+ l 1 1+{Σn i l l b˜− l 1
i=1 wi ( 1+t− ) }
1+{Σn 1+{Σn
i=1 wi ( ) }l
l
i 1+t−
b̃+ −

−1 −1 −1
1+µ− 1
≤ 1+µ− 1
≤ 1+µ− 1
,
b̃− )l } l 1+{Σn i )l } l b̃+ )l } l
1+{Σn w ( i=1 wi ( −µ−
1+{Σn w (
−µ− −µ−
i=1 i i=1 i
i
b̃− b̃+
−1 −1 −1
1+ν − 1
≤ 1+ν − 1
≤ 1+ν − 1
.
b̃− )l } l
i=1 wi ( −ν − ) }
1+{Σn
1+{Σn w (
i l l
1+{Σn w ( b̃+ )l } l
−ν − −ν −
i=1 i i=1 i
i
b̃− b̃+

Hence, BNDWGA(b̃− , b̃− , ..., b̃− ) ≤BNDWGA(b̃1 , b̃2 , ..., b̃n ) ≤ BNDWGA(b̃+ , b̃+ , ..., b̃+ )
holds.

Example 4.3. Let b˜1 =< 0.6, 0.7, 0.3, −0.6, −0.3, −0.5 >, b˜2 =< 0.5, 0.4, 0.6, −0.6, −0.7, −0.3 >,
b˜3 =< 0.6, 0.7, 0.4, −0.9, −0.7, −0.7 > and b˜4 =< 0.2, 0.6, 0.8, −0.6, −0.3, −0.9 > be four
BNNs and let the weight vector of BNNs b˜i (i = 1, 2, 3, 4) be w = ( 21 , 41 , 18 , 18 )T . w1 = 12 ,
w2 = 14 , w3 = 18 and w4 = 18 are the weights of b̃i (i = 1, 2, 3, 4) such that Σi=1 wi = 1.

12
Then, by theorem (4.2), for l = 3


 ⟨ 1
1 ,
 1+{ 21 ( 1−0.6
 )3 + 14 ( 1−0.5 )3 + 18 ( 1−0.6 )3 + 18 ( 1−0.2 )3 } 3


0.6 0.5 0.6 0.2

 1− 1
1 ,

 1 0.7 3 + 1 ( 0.4 )3 + 1 ( 0.7 )3 + 1 ( 0.6 )3 } 3


1+{ 2
( 1−0.7
) 4 1−0.4 8 1−0.7 8 1−0.6

1 − 1
1 ,
1+{ 2 ( 1−0.3 ) + 4 ( 1−0.6 ) + 8 ( 1−0.4 ) + 8 ( 1−0.8 ) } 3
1 0.3 3 1 0.6 3 1 0.4 3 1 0.8 3
BN DW GA(b̃1 , b̃2 , b̃3 , b̃4 ) =


1
1 − 1,

 1+{ 12 ( 1−0.6
0.6
)3 + 14 ( 1+0.6
0.6
)3 + 18 ( 1−0.9
0.9
)3 + 81 ( 1−0.6
0.6
)3 } 3



 −1
,

 1 1−0.3 3 1 1−0.7 3 1 1−0.7 3 1 1−0.3 3 1


1+{ ( ) + ( ) + 8 ( 0.7 ) + 8 ( 0.3 ) } 3

2 0.3 4 0.7
 −1
1 ⟩,
1+{ 12 ( 1−0.5
0.5
)3 + 14 ( 1−0.3
0.3
)3 + 18 ( 1−0.7
0.7
)3 + 18 ( 1−0.9
0.9
)3 } 3

BN DW GA(b̃1 , b̃2 , b̃3 , b̃4 ) = ⟨0.3294, 0.6706, 0.6747, −0.8198, −0.3336, −0.393⟩.

4.2 Bipolar neutrosophic Dombi ordered weighted geometric


aggregation operator
+ − − −
Definition 4.4. Let b˜i =< t+ +
i , µi , νi , ti , µi , νi > (i = 1, , 2, 3, ..., n) be a family of
BNNs. A mapping BNDOWGA: U n → U is called bipolar neutrosophic Dombi ordered
weighted geometric aggregation (BNDOWGA) operator, if it satisfies

n ∧wj ∧w1 ∧w2 ∧wn
BN DOW GA(b̃1 , b̃2 , ..., b̃n ) = b̃σ(j) = b̃σ(1)
D
⊗D b̃σ(2)
D
⊗D ... ⊗D b̃σ(n)
D
(8)
j=1

where σ is permutation that orders the elements: b̃σ(1) ≥ b̃σ(2) ≥ ... ≥ b̃σ(n) . where
w = (w1 , w2 , ..., wn )T is the weight vector of b̃σ(j) (j = 1, 2, ..., n), wj ∈ [0, 1] and
Σj=1 wj = 1.

+ − − −
Theorem 4.5. Let b˜i =< t+ +
i , µi , νi , ti , µi , νi > (i = 1, 2, 3, ..., n) be a family of BNNs
and w = (w1 , w2 , ..., wn ) be the weight vector of bi , wi ∈ [0, 1] and Σi=1 wi = 1. Then,
T

the aggregated value of BNDOWGA operator is still a BNN, which is calculated by the
following formula
∧w1 ∧w2 ∧wn
BNDOWGA(b̃1 , b̃2 , ..., b̃n ) = b̃σ(1)
D
⊗D b̃σ(2)
D
⊗D ... ⊗D b̃σ(n) D



 ⟨ 1
,1 − 1
,1 − 1
,
 +
 1+{Σnj=1 wj ( 1−t+j )l } 1l n
1+{Σj=1 wj (
µ+
j
)}
l
1
l
ν+ 1
1+{Σj=1 wj ( j + )l } l
n
t 1−µ + 1−ν
= j j j
(9)


1
− 1, −1
, −1
⟩.
 1+{Σn w ( −tj )l } 1l


1+{Σn w (
1+µ −
j l 1
) } l 1+{Σn w (
1+ν −
j l 1
) }l
1+t− −µ− −ν −
j=1 j j=1 j j=1 j
j j j

13
By the mathematical induction, we can prove theorem (4.5).

Proof: If n = 2 based of the Dombi operations on BNNs in definition (4.4), we can


obtain the following result: ∧w1 ∧w2
BNDOWGA (b̃1 , b̃2 ) = b̃σ(1) D
⊗D b̃σ(2)
D


⟨
 ,1 − ,1 −
1 1 1
,
 1−t+ 1−t+
1+{w1 ( + )l +w2 ( + )l } l
1 2
1
1+{w1 (
µ+1 )l +w2 (
µ+
2
+) }
l l
1
1+{w1 (
+
ν1 +
ν2
+ ) +w2 ( 1−ν + ) }
l
1
l l
t1 t2 1−µ+ 1−µ 1−ν
= 1 2 1 2


1
− 1, −1
, −1
⟩,
 1+{w1 ( −t−1 )l +w2 ( −t−2 )l } 1l 1+{w1 (
1+µ− − 1
1 )l +w ( 1+µ2 )l } l 1+{w1 (
1+ν1−
)l +w2 (

1+ν2 1
)l } l
− − − 2 − − −
 1+t1 1+t2 −µ1 −µ2 −ν1 −ν2


 ⟨ 1
,1 − 1
,1 − 1
,
 1+{Σ2 w ( 1−t+j )l } 1l
 µ+ 1
1+{Σ2j=1 wj ( j + )l } l
ν+ 1
1+{Σ2j=1 wj ( j + )l } l
j=1 j t+ 1−µ 1−ν
= j j j


1
− 1, −1
, −1
⟩.
 −
 1+{Σ2 w ( −tj )l } 1l 1+{Σ2 w (
1+µ−
j l 1
) } l 1+{Σ2 w (
1+ν −
j l 1
) }l
1+t− −µ− −ν −
j=1 j j=1 j j=1 j
j j j

If n = m, based on equation (9), then we have the following equation:


⊗ ∧wj
BNDOWGA(b̃1 , b̃2 , ..., b̃m ) = m j=1 σ(j) =



 ⟨ 1
,1 − 1
,1 − 1
,
 1−t+
 1+{Σj=1 wj ( + ) } l
j 1 µ+ 1
1+{Σj=1 wj ( j + )l } l
ν+
j 1
j=1 wj ( 1−ν + ) }
m l m 1+{Σm l l
t 1−µ
j j j


1
− 1, , −1 −1
⟩.
 m
−t−
 1+{Σj=1 wj ( − )l } l
j 1
m
1+{Σj=1 wj (
1+µ−
j
)
1
l} l 1+{Σm w (
1+ν −
j l 1
) } l
1+t − j=1
−µ
j − −ν
j j j
If n = m + 1, then there is following result:
⊗ ∧w j ⊗ ∧w
m+1
BNDOWGA(b̃1 , b̃2 , ..., b̃m , b̃m+1 ) = m b̃ b̃σ(m+1)
 j=1 σ(j)

 ⟨ 1
,1 − 1
,1 − 1
,

 1+{Σm 1−t+
j 1 µ+
j 1 ν+
j 1
j=1 wj ( t+ ) } 1+{Σmj=1 wj ( 1−µ+ ) } 1+{Σm j=1 wj ( 1−ν + ) }
l l l l l l

= j j j


1
− 1, 1
, 1


 1+{Σm −t−
j 1 1+µ−
j l 1
1+ν −
j l 1
j=1 wj ( 1+t− ) } 1+{Σmj=1 wj ( −µ− ) } 1+{Σm j=1 wj ( −ν − ) }
l l l l
 j j j


 ⟨ 1
, 1 − 1
, 1 − 1
,
 1−t+
 1+{wm+1 ( + ) } l m+1 l
1
1+{wm+1 (
µ+m+1
) l }
1
l 1+{w (
+
νm+1
)
1
l} l
m+1
1−µ+ +
⊗D
tm+1 m+1 1−νm+1



1
− 1, 1
, 1


 1+{wm+1 ( −tm+1

)
1
l} l 1+{w (
1+µ− m+1 l 1
) } l 1+{w (

1+νm+1
)
1
l} l
1+t− −µ− −
m+1 m+1
 m+1 m+1 −νm+1


 ⟨ 1
,1 − 1
,1 − 1
,
 m+1
1−t+
 1+{Σj=1 wj ( + )l } l j 1 m+1
1+{Σj=1 wj (
µ+
j
)l } l
1 ν+
1+{Σj=1 wj ( j + )l } l
m+1 1
t 1−µ + 1−ν
= j j j


1
− 1, −1
, −1
⟩.
 −
 1+{Σm+1 w ( −tj )l } 1l 1+{Σm+1 w (
1+µ−
j l 1
) } l 1+{Σm+1 w (
1+ν −
j l 1
) }l
1+t− −µ− −ν −
j=1 j j=1 j j=1 j
j j j
Hence, Theorem (4.5) is true for n = m + 1. Thus, equation (9) holds for all n.

The BNDOWGA operator also contains the following properties:

14
+ − − −
(1) Idomopotency: Let all the BNNs be b˜i =< t+ +
i , µi , νi , ti , µi , νi >= b̃(i = 1, 2, 3, ..., n),
where w = (w1 , w2 , ..., wn )T be the weight vector of b̃i , wi ∈ [0, 1] and Σi=1 wi = 1.
Then, BNDOWGA(b̃1 , b̃2 , ..., b̃n ) = b̃.

(2) Monotonicity: Let b̃i (i = 1, 2, ..., n) and b̃,i (i = 1, 2, ..., n) be two families of BNNs,
where w = (w1 , w2 , ..., wn )T be the weight vector of b̃i and b̃,i , wi ∈ [0, 1] and Σi=1 wi =
1. For all i ,if b̃i ≥ b̃,i , then BNDOWGA(b̃1 , b̃2 , ..., b̃n ) ≥ BNDOWGA(b̃,1 , b̃,2 , ..., b̃,n ).
+ − − −
(3) Boundedness: Let b˜i =< t+ +
i , µi , νi , ti , µi , νi > (i = 1, 2, 3, ..., n) be a family of
BNNs, where w = (w1 , w2 , ..., wn )T be the weight vector of bi , wi ∈ [1, 0] and
Σi=1 wi = 1. Therefore, we have
BNDOWGA(b̃− , b̃− , ..., b̃− ) ≤ BNDOWGA(b̃1 , b̃2 , ..., b̃n ) ≤ BNDOWGA(b̃+ , b̃+ , ..., b̃+ ),
where b̃− =< t+ , µ+ , νb̃+− , t− , µ− , νb̃−− >
{ b̃− b̃− b̃− b̃−
< min(t+ + + + + + + +
1 , t2 , ...tn , ), max(µ1 , µ2 , ...µn ), max(ν1 , ν2 , ..., νn ),
+
=
max(t− − − − − − − −
1 , t2 , ...tn ), min(µi , µi , ...µi ), min(ν1 , ν2 , ..., νn ) >

and
b̃+ =< t+ , µ+ , νb̃++ , t− , µ− , νb̃−+ >
{ b̃+ b̃+ b̃+ b̃+
< max(t+ + + + + + + +
1 , t2 , ...tn , ), min(µ1 , µ2 , ...µn ), min(ν1 , ν2 , ..., νn ),
+
=
min(t− − − − − − − −
1 , t2 , ...tn ), max(µi , µi , ...µi ), max(ν1 , ν2 , ..., νn ) > .

+ − − −
Proof (1) Since b˜i =< t+ +
i , µi , νi , ti , µi , νi >= b̃(i = 1, , 2, 3, ..., n). Then, by using
equation (9), we can obtain the following result:


 ⟨ 1
,1 − 1
,1 − 1
,
 +
 1+{Σnj=1 wj ( 1−t+j )l } 1l µ+
j 1 ν+
j 1
j=1 wj ( 1−µ+ ) }
1+{Σn 1+{Σn )l } l
l l
t j=1 wj ( 1−ν +
BNDOWGA(b̃1 , b̃2 , ..., b̃n ) = j j j


1
− 1, −1
, −1

 −
 1+{Σnj=1 wj ( −tj− )l } 1l 1+µ−
j l 1
1+ν −
j l 1
− ) } − ) }
n
1+{Σj=1 wj ( l n
1+{Σj=1 wj ( l
1+t −µ −ν
j j j



⟨ 1+{( 1−t+ )l } 1l , 1 −
1 1
µ+ 1 ,1 − ν+
1
1 ,
1+{( )l } l 1+{( )l } l
= t+ 1−µ+ 1−ν +
=< t+ , µ+ , ν + , t− , µ− , ν − >= b̃.
 −1 −1
 1+{( −t− )l } 1l − 1, 1+{( 1+µ− )l } 1l , 1+{( 1+ν − )l } 1l ⟩
1

1+t− −µ− −ν −

Hence, BNDOWGA(b̃1 , b̃2 , ..., b̃n ) = b̃ holds.

(2) based on the equation (9), the property is obvious.

(3) Let b̃− =< t+b̃−


, µ+
b̃−
, νb̃+− , t−
b̃−
, µ−
b̃−
, νb̃−− > and b̃+ =< t+
b̃+
, µ+
b̃+
, νb̃++ , t−
b̃+
, µ−
b̃+
, νb̃−+ >. There
are the following inequalities:
1
1−t+ 1
≤ 1
1−t+ 1
≤ 1
1−t+
,
1+{Σn b̃− )l } l 1+{Σn j
)l } l + l 1

j=1 wj ( 1+
j=1 wj ( t+
1+{Σn
j=1 wj ( ) }l
t+
b̃− j
b̃+

15
1− 1
µ+ 1
≤1− 1
µ+ 1
≤1− µ+
1
1
,
b̃+ j b̃−
j=1 wj ( 1−µ+ ) } j=1 wj ( 1−µ+ ) }
1+{Σn j=1 wj ( 1−µ+ ) }
1+{Σn l l l l 1+{Σn l l

b̃+ j −

1− 1
ν+ 1
≤1− 1
ν+ 1
≤1− ν+
1
1
,
b̃+ j b̃−
j=1 wj ( 1−ν + ) } j=1 wj ( 1−ν + ) }
1+{Σn j=1 wj ( 1−ν + ) }
1+{Σn l l l l 1+{Σn l l
j
b−
b̃+ ˜
1
−t−
−1≤ 1
−t− 1
−1≤ 1
−t−
− 1,
b˜+ l 1 j
b˜− l 1
j=1 wj ( 1+t− ) }
1+{Σn l l
1+{Σn w ( ) }l 1+{Σn
j=1 wj ( ) }l
1+t− 1+t−
j=1 j
j
b̃+ −

−1 −1 −1
1+µ−
≤ 1+µ−
≤ 1+µ−
,
1
b̃− l l j l 1 1
b̃+ )l } l
j=1 wj ( −µ− ) }
1+{Σn j=1 wj ( −µ− ) }
1+{Σn 1+{Σn
j=1 wj (
l
−µ−
b̃ − j b̃ +
−1 −1 −1
1+ν −
≤ 1+ν −
≤ 1+ν −
.
1
b̃− l l j l 1 1
b̃+ )l } l
j=1 wj ( −ν − ) }
1+{Σn j=1 wj ( −ν − ) }
1+{Σn 1+{Σn
j=1 wj (
l
−ν −
b̃− j b̃+

Hence, BNDOWGA(b̃− , b̃− , ..., b̃− ) ≤BNDOWGA(b̃1 , b̃2 , ..., b̃n ) ≤ BNDOWGA(b̃+ , b̃+ , ..., b̃+ )
holds.

Example 4.6. Let b˜1 =< 0.6, 0.7, 0.3, −0.6, −0.3, −0.5 >, b˜2 =< 0.5, 0.4, 0.6, −0.6, −0.7, −0.3 >,
b˜3 =< 0.6, 0.7, 0.4, −0.9, −0.7, −0.7 > and b˜4 =< 0.2, 0.6, 0.8, −0.6, −0.3, −0.9 > be four
BNNs and let the weight vector of BNNs b˜i (i = 1, 2, 3, 4) be w = ( 21 , 41 , 18 , 18 )T . w1 = 12 ,
w2 = 14 , w3 = 18 and w4 = 18 are the weights of b̃i (i = 1, 2, 3, 4) such that Σi=1 wi = 1.
Then, by theorem (4.5), for l = 3


 ⟨ 1 1−0.6 3 1 1−0.5 31 1 1−0.6 3 1 1−0.2 3 1 ,

 1+{ 2 ( 0.6 ) + 4 ( 0.5 ) + 8 ( 0.6 ) + 8 ( 0.2 ) } 3

1 −


1
1 ,

 1+{ 12 ( 1−0.7
0.7
)3 + 14 ( 1−0.4
0.4
)3 + 81 ( 1−0.7
0.7
)3 + 81 ( 1−0.6
0.6
)3 } 3



1 − 1
1 ,
1+{ 2 ( 1−0.4 ) + 4 ( 1−0.6 ) + 81 ( 1−0.3
1 0.4 3 1 0.6 3 0.3
)3 + 81 ( 1−0.8
0.8
)3 } 3
BN DOW GA(b̃1 , b̃2 , b̃3 , b̃4 ) =


1
1 − 1,

 1 0.9 3 + 1 ( 0.6 )3 + 1 ( 0.6 )3 + 1 ( 0.6 )3 } 3


1+{ (
2 1+0.9
) 4 1+0.6 8 1+0.6 8 1+0.6

 −1
,

 1 1−0.7 3 1 1−0.7 3 1 1−0.3 3 1 1−0.3 3 1


1+{ ( ) + ( ) + 8 ( 0.3 ) + 8 ( 0.3 ) } 3

2 0.7 4 0.7
 −1
⟩,
1 1−0.7 3 1 1−0.3 3 1 1−0.5 3 1 1−0.9 3 1
1+{ 2 ( 0.7
) +4( 0.3
) +8( 0.5
) +8( 0.9
) }3

BN DOW GA(b̃1 , b̃2 , b̃3 , b̃4 ) = ⟨0.3294, 0.6706, 0.6753, −0.8774, −0.4034, −0.4008⟩.

5 Bipolar Neutrosophic Dombi prioritized aggrega-


tion operator
The prioritized average operator (PA) was defined by Yager (2008), which was given as
the following definition.

16
Definition 5.1. Let C = {C1 , C2 , ..., Cn } be a collection of attributes and have a pri-
oritization between the attributes followed by the linear ordering C1 ≻ C2 ≻ ... ≻ Cn ,
which imply that Cτ has a higher priority than Cρ , if τ < ρ. The value of Cρ (x) is the
performance of any alternative x under attribute Cτ (x), which satisfies Cτ ∈ [0, 1], If

P A(Cτ (x)) = Σnρ=1 υρ Cρ (x) (10)

where υρ = Σn~ρ ~ρ ; ~ρ = Πρ−1


τ =1 Cτ (x)(ρ = 2, 3, ..., n), ~ = 1. Then, PA is called
ρ=1
the prioritized averaging operator.

5.1 Bipolar neutrosophic Dombi prioritized weighted geometric


aggregation operator
+ − − −
Definition 5.2. Let b˜i =< t+ +
i , µi , νi , ti , µi , νi > (i = 1, , 2, 3, ..., n) be a family of
BNNs. A mapping BNDPWGA: U → U is called bipolar neutrosophic Dombi prioritized
n

weighted geometric aggregation (BNDPWGA) operator, if it satisfies



n ∧υ
i
∧υ1 ∧υ 2 ∧υ n
BN DP W GA(b̃1 , b̃2 , ..., b̃n ) = b̃i = b̃1 D ⊗D b̃2 D ⊗D ... ⊗D b̃n D (11)
i=1
~i wi
∏i−1
where υi = , ~i = s(b̃k ) (i = 2, 3, ..., n), ~1 = 1 and s(b̃k ) is the value
i=1 ~i wi
Σn k=1
of score for b̃i . where w = (w1 , w2 , ..., wn )T is the weight vector of b̃i (i = 1, 2, ..., n),
wi ∈ [0, 1] and Σi=1 wi = 1.

+ − − −
Theorem 5.3. Let b˜i =< t+ +
i , µi , νi , ti , µi , νi > (i = 1, 2, 3, ..., n) be a family of BNNs
and w = (w1 , w2 , ..., wn )T be the weight vector of bi , wi ∈ [0, 1] and Σi=1 wi = 1. Then,
the aggregated value of BNDPWGA operator is still a BNN, which is calculated by the
following formula
∧υ1 ∧υ2 ∧υn
BNDPWGA(b̃1 , b̃2 , ..., b̃n ) = b̃1 ⊗D b̃2 ⊗D ... ⊗D b̃n D
D D



 ⟨ 1
,1 − 1
,1 − 1
,
 1+{Σni=1 Σn~i w~i w ( 1−t+i )l } 1l
+ µ+ 1 ν+ 1
~i wi ~i wi
n
1+{Σi=1 Σn ~ w ( i
+ ) }l
l 1+{Σi=1 Σn ~ w ( i + )l } l
n
i=1 i i t j=1 i i 1−µi i=1 i i 1−νi
= i
 −1 −1
 1+{Σn n~i wi ( −t−i )l } 1l − 1, 1+{Σn n~i wi ( 1+µ−i )l } 1l , 1+{Σn n~i wi ( 1+νi− )l } 1l ⟩
1

i=1 Σ ~ w
i=1 i i 1+t− i=1 Σ ~ w
i=1 i i −µ− i=1 Σ ~ w
i=1 i i −ν −
i i i
(12)
By the mathematical induction, we can prove theorem (5.3).

Proof: If n = 2 based of the Dombi operations on BNNs in definition (5.2), we can


obtain the following result:
∧υ1 ∧υ2
BNDPWGA (b̃1 , b̃2 ) = b̃1 ⊗D b̃2 D
D

17


 ⟨ 1
,1 − 1
,


~1 w1 1−t+1 )l + ~2 w2 1−t+ 1
2 )l } l ~1 w1 µ+
1 )l + ~2 w2 µ+
( 2 + )l } l
1


1+{ ~1 w1 +~ 2 w 2
( +
t1 ~1 w 1 +~ 2 w 2
( +
t2
1+{ ~1 w 1 +~ 2 w2
(
1−µ1+ ~ 1 w 1 +~ 2 w 2 1−µ2

1− +
1
+ , −t−
1
−t−
− 1,
= ~1 w1 ν1 ~2 w2 ν2 1
l} l ~1 w1 ~2 w2 1
2 )l } l
 1+{ ( ) l+ ( ) 1+{ ( 1 )l + (


~1 w1 +~2 w2 1−ν + ~1 w1 +~2 w2 1−ν + ~1 w1 +~2 w2 1+t − ~1 w1 +~2 w2 1+t −


1 2 1 2
 −1 −1
 1+{ ~1 w1 ( 1+µ1 )l + ~2 w2 ( 1+µ2 )l } 1l 1+{ ~1 w1 ( 1+ν1 )l + ~2 w2 ( 1+ν2 )l } 1l ⟩,
 − − , − −
~1 w1 +~2 w2 −µ− ~1 w1 +~2 w2 −µ− ~1 w1 +~2 w2 −ν − ~1 w1 +~2 w2 −ν −
 1 2 1 2


 ⟨ 1
,1 − 1
,1 − 1
,
 1+{Σ2i=1 2~i wi ( 1−t+i )l } 1l
+ µ+ 1 ν+ 1
~ w ~ w
1+{Σ2i=1 2 i i ( i + )l } l 1+{Σ2i=1 2 i i (Σ2i=1 i + )l } l
Σ ~i wi t Σ ~ w Σ ~ w
i=1 i i 1−µi i=1 i i 1−ν
= i=1 i i


1
− 1, −1
, −1
⟩.
 1+{Σ2 ~i wi −t−
( i )l } l
1
1+{Σ2
~i wi
(
1+µ− 1
i )l } l 1+{Σ2
~i wi
(
1+ν − 1
i )l } l
i=1 Σ2 ~ w
i=1 i i 1+t− i=1 Σ2 ~ w
i=1 i i −µ− i=1 Σ2 ~ w
i=1 i i −ν −
i i i

If n = m, based on equation (12), then we have the following equation:


⊗ ∧υ
i
BNDPWGA(b̃1 , b̃2 , ..., b̃m ) = m b̃
 i=1 i

⟨ 1
,1 − 1
,1 − 1
,
 1+{Σm ~i wi 1−t+
i=1 Σm ~i wi ( t+ ) }
i l
1
l m ~i wi
1+{Σi=1 Σm ~ w (
µ+
i
+ )}
l
1
l
~i wi ν+
1+{Σi=1 Σm ~ w ( i + )l } l
m
1

i=1 i=1 i i 1−µi i=1 i i 1−νi


= i
 −1 −1
 1+{Σm m~i wi ( −t−i )l } 1l − 1, 1+{Σm m~i wi ( 1+µ−i )l } 1l , 1+{Σm m~i wi ( 1+νi− )l } 1l ⟩.
1

i=1 Σ ~ w
i=1 i i 1+t− i=1 Σ ~ w
i=1 i i −µ− i=1 Σ ~ w
i=1 i i −ν −
i i i
If n = m + 1, then there is following result:
⊗ ∧υ ⊗ ∧υ
i m+1
BNDPWGA(b̃1 , b̃2 , ..., b̃m , b̃m+1 ) = m b̃i b̃m+1
 i=1
⟨

1
,1 − 1
,1 − 1
,
 1+{Σm ~i wi
i=1 Σm ~i wi (
1−t+
+
1
i )l } l 1+{Σ m
i=1 Σ
~i wi
m ~ w (
µ+
i
+ )l} l
1
1+{Σ m
i=1 Σ
~i wi
m ~ w (
ν+
i
+
1
)l } l
i=1 t i=1 i i 1−µi i=1 i i 1−νi
= i


1
− 1, 1
, 1

 1+{Σm
~i wi
(
−t−
i )l } l
1
1+{Σ m ~i wi
(
1+µ− 1
i )l } l 1+{Σ m ~i wi
(
1+ν −
i )l } l
1
m ~ w − m ~ w − m ~ w −

i=1 Σ i=1 Σ i=1 Σ
i=1 i i 1+ti i=1 i i −µi i=1 i i −νi


 ⟨ ~ w 1
,1 − 1
,1 − 1
,

 1+{ m+1m+1 m+1 1−t+
( + m+1
) }l
l
1 ~m+1 wm+1
1+{ m+1 (
µ+
m+1
) l }
1
l 1+{
~m+1 wm+1
(
+
νm+1
)l} l
1
~ w ~ w 1−µ+ Σm+1 ~i wi 1−νm+1 +
⊗D
Σ tm+1 Σ
i=1 i i i=1 i i m+1 i=1


1
− 1, 1
, 1


 1+{ ~Σm+1 wm+1 −t−
i )l } l
1 ~
1+{ m+1
wm+1 1+µm+1 l 1

}
~m+1 wm+1 1+νm+1 l 1

}
m+1 ~ w ( 1+t− ( ) l 1+{ ( ) l
Σm+1 ~i wi −µ− Σm+1 ~i wi −
−νm+1
 i=1 i i m+1 i=1 m+1 i=1


 ⟨ 1
,1 − 1
,1 − 1
,
 1+{Σi=1 m+1
m+1 ~ w
i i
1−t+
( +i )l } l
1 m+1
1+{Σi=1 m+1
~ w
i i
µ+
( i )l } l
1 m+1
1+{Σi=1 m+1
~ w
i i
ν+
( i )l } l
1
Σ ~i wi t Σ ~i wi 1−µ+ Σ ~i wi 1−ν +
= i=1 i i=1 i i=1 i


1
− 1, 1
, 1

 1+{Σm+1 ~i wi ( −t−i )l } 1l 1+{Σm+1
~i wi
(
1+µ−
i )l } l
1
1+{Σm+1
~i wi
(
1+ν −
i )l } l
1
i=1 Σm+1 ~ w 1+t− i=1 Σm+1 ~ w −µ− i=1 Σm+1 ~ w −ν −
i=1 i i i i=1 i i i i=1 i i i
Hence, Theorem (5.3) is true for n = m + 1. Thus, equation (12) holds for all n.

The BNDPWGA operator also contains the following properties:

+ − − −
(1) Idomopotency: Let all the BNNs be b˜i =< t+ +
i , µi , νi , ti , µi , νi >= b̃ (i = 1, 2, 3, ..., n),
where w = (w1 , w2 , ..., wn )T be the weight vector of b̃i , wi ∈ [0, 1] and Σi=1 wi = 1.
Then, BNDPWGA(b̃1 , b̃2 , ..., b̃n ) = b̃.
(2) Monotonicity: Let b̃i (i = 1, 2, ..., n) and b̃,i (i = 1, 2, ..., n) be two families of BNNs,

18
where w = (w1 , w2 , ..., wn )T be the weight vector of b̃i and b̃,i , wi ∈ [0, 1] and Σi=1 wi =
1. For all i ,if b̃i ≥ b̃,i , then BNDPWGA(b̃1 , b̃2 , ..., b̃n ) ≥ BNDPWGA(b̃,1 , b̃,2 , ..., b̃,n ).
+ − − −
(3) Boundedness: Let b˜i =< t+ +
i , µi , νi , ti , µi , νi > (i = 1, , 2, 3, ..., n) be a family
of BNNs, where w = (w1 , w2 , ..., wn )T be the weight vector of bi , wi ∈ [0, 1] and
Σi=1 wi = 1. Therefore, we have
BNDPWGA(b̃− , b̃− , ..., b̃− ) ≤ BNDPWGA(b̃1 , b̃2 , ..., b̃n ) ≤ BNDPWGA(b̃+ , b̃+ , ..., b̃+ ),
where b̃− =< t+ , µ+ , νb̃+− , t− , µ− , νb̃−− >
{ b̃− b̃− b̃− b̃−
< min(t+ + + + + + + +
1 , t2 , ...tn , ), max(µ1 , µ2 , ...µn ), max(ν1 , ν2 , ..., νn ),
+
=
max(t− − − − − − − −
1 , t2 , ...tn ), min(µi , µi , ...µi ), min(ν1 , ν2 , ..., νn ) >

and
b̃+ =< t+ , µ+ , νb̃++ , t− , µ− , νb̃−+ >
{ b̃+ b̃+ b̃+ b̃+
< max(t+ + + + + + + +
1 , t2 , ...tn , ), min(µ1 , µ2 , ...µn ), min(ν1 , ν2 , ..., νn ),
+
=
min(t− − − − − − − −
1 , t2 , ...tn ), max(µi , µi , ...µi ), max(ν1 , ν2 , ..., νn ) > .

+ − − −
Proof (1) Since b˜i =< t+ +
i , µi , νi , ti , µi , νi >= b̃(i = 1, , 2, 3, ..., n). Then, by using
equation (12), we can obtain the following result:

⊗ ∧υ
i
BNDPWGA(b̃1 , b̃2 , ..., b̃n ) = ni=1 b̃i


 ⟨ 1
,1 − 1
,1 − 1
,
 1+{Σni=1 Σn~i w~i w ( 1−t+i )l } 1l
+ µ+ 1 ν+ 1
~i wi l} l ~i wi
1+{Σn
i=1 Σ n ~ w ( i
+ ) 1+{Σn
i=1 Σ n ~ w (
i
+ )l } l
i=1 i i t i=1 i i 1−µi i=1 i i 1−νi
= i
 −1 −1
 1+{Σn n~i wi ( −t−i )l } 1l − 1, 1+{Σn n~i wi ( 1+µ−i )l } 1l , 1+{Σn n~i wi ( 1+νi− )l } 1l ⟩
1

i=1 Σ ~ w
i=1 i i 1+t− i=1 Σ ~ w
i=1 i i −µ− i=1 Σ ~ w
i=1 i i −ν −
i i i



⟨ 1+{( 1−t+ )l } 1l , 1 −
1 1
µ+ 1 ,1 − ν+
1
1 ,
1+{( )l } l 1+{( )l } l
= t+ 1−µ+ 1−ν +
=< t+ , µ+ , ν + , t− , µ− , ν − >= b̃.
 −1 −1
 1+{( −t− )l } 1l − 1, 1+{( 1+µ− )l } 1l , 1+{( 1+ν − )l } 1l ⟩
1

1+t− −µ− −ν −

Hence, BNDPWGA(b̃1 , b̃2 , ..., b̃n ) = b̃ holds.

(2) based on the equation (12), the property is obvious.

(3) Let b̃− =< t+b̃−


, µ+
b̃−
, νb̃+− , t−
b̃−
, µ−
b̃−
, νb̃−− > and b̃+ =< t+
b̃+
, µ+
b̃+
, νb̃++ , t−
b̃+
, µ−
b̃+
, νb̃−+ >. There
are the following inequalities:
1
1−t+ 1
≤ 1
~ w 1−t+ 1
≤ 1
1−t+
,
~i wi − 1+{Σn i i ( +i )l } l ~i wi ˜ 1
1+{Σn
i=1 Σn ~ w
( +b̃ )l } l i=1 Σn ~ w 1+{Σn
+
( +b )l } l
i=1 i i 1 i=1 i i t i=1 Σn ~ w
− i i=1 i i t
b̃ b̃+
1− 1
µ+ 1
≤ 1− 1
~i wi µ+ 1
≤ 1− 1
µ+ 1
,
~i wi b̃+ )l } l 1+{Σn ( i )l } l ~i wi b̃− )l } l
1+{Σn
i=1 Σ n ~ w
( + i=1 Σn ~i wi 1−µ+ 1+{Σn
i=1 Σn ~ w
( +
i=1 i i 1−µ + i=1 i i=1 i i 1−µ ˜
b̃ b−
1− 1
ν+ 1
≤ 1− ~i wi
1
ν+ 1
≤ 1− 1
ν+ 1
,
~i wi ~i wi b̃−
i=1 Σn ~i wi ( 1−ν + ) }
1+{Σn
b̃+ i l l
i=1 Σn ~i wi ( 1−ν + ) }
1+{Σn i=1 Σn ~i wi ( 1−ν + ) }
1+{Σn
l l l l
i=1 i=1 i i=1
b−
b̃ + ˜

19
1
−t−
−1 ≤ 1
~i wi −t− 1
−1 ≤ 1
−t−

1+{Σn
~i wi
( b˜+ l 1
) }l 1+{Σn
i=1 Σn ( i )l } l
~i wi 1+t− 1+{Σn
~i wi
( b˜− l 1
) }l
i=1 Σn ~i wi 1+t− i=1 i i=1 Σ n ~i wi 1+t−
i=1 i=1
b̃+ −

1,
−1 −1 −1
1+µ− 1
≤ ~i wi 1+µ− 1
≤ 1+µ− 1
,
~i wi b̃− l l ~i wi
i=1 Σn ~i wi ( −µ− ) }
1+{Σn i l l b̃+ l l
i=1 Σn ~i wi ( −µ− ) }
1+{Σn
i=1 i=1 Σn ~i wi ( −µ− ) }
1+{Σn
i=1 i i=1
b̃− b̃+
−1 −1 −1
1+ν − 1
≤ ~i wi 1+ν − 1
≤ 1+ν − 1
.
~i wi b̃− l l 1+{Σn i )l } l ~i wi b̃+ l l
i=1 Σn ~i wi ( −ν − ) } i=1 Σn ~i wi ( −ν − ) }
1+{Σn i=1 Σn (
~i wi −ν − 1+{Σn
i=1 i=1 i i=1
b̃− b̃+

Hence, BNDPWGA(b̃− , b̃− , ..., b̃− ) ≤ BNDPWGA(b̃1 , b̃2 , ..., b̃n ) ≤ BNDPWGA(b̃+ , b̃+ , ..., b̃+ )
holds.

Example 5.4. Let b˜1 =< 0.6, 0.7, 0.3, −0.6, −0.3, −0.5 >, b˜2 =< 0.5, 0.4, 0.6, −0.6, −0.7, −0.3 >,
b˜3 =< 0.6, 0.7, 0.4, −0.9, −0.7, −0.7 > and b˜4 =< 0.2, 0.6, 0.8, −0.6, −0.3, −0.9 > be
four BNNs and let the weight vector of BNNs b˜i (i = 1, 2, 3, 4) be w = ( 21 , 14 , 18 , 18 )T .
Now, by definition (10), ~1 = 1, ~2 = s(b̃1 ) = 0.4667, ~3 = s(b̃1 )s(b̃2 ) = 0.2256
and ~4 = s(b̃1 )s(b̃2 )s(b̃3 ) = 0.1128. w1 = 21 , w2 = 14 , w3 = 18 and w4 = 81 are the
weights of b̃i (i = 1, 2, 3, 4) such that Σi=1 wi = 1. By definition (10), Σ4~1 w~1i wi = 0.7588,
i=1
~2 w2 ~3 w3 ~4 w4
Σ4i=1 ~i wi
= 0.1771, Σ4i=1 ~i wi
= 0.0428, Σ4i=1 ~i wi
= 0.0214. Then, by theorem (5.3), for l = 3


 ⟨ 1
1 ,
 1+{0.7588( 1−0.6
 )3 +0.1771( 1−0.5 )3 +0.0428( 1−0.6 )3 +0.0214( 1−0.2 )3 } 3


0.6 0.5 0.6 0.2

 1− 1
1 ,

 0.7 3 +0.1771( 0.4 )3 +0.0428( 0.7 )3 +0.0214( 0.6 )3 } 3


1+{0.7588( 1−0.7
) 1−0.4 1−0.7 1−0.6

1 − 1
1 ,
1+{0.7588( 1−0.3 ) +0.1771( 1−0.6 ) +0.0428( 1−0.4 ) +0.0214( 1−0.8 ) } 3
0.3 3 0.6 3 0.4 3 0.8 3
BN DP W GA(b̃1 , b̃2 , b̃3 , b̃4 ) =


1
1 − 1,

 0.6 0.6
)3 +0.1771( 1−0.6 0.9
)3 +0.0428( 1−0.9 0.6
)3 +0.0214( 1−0.6 )3 } 3


1+{0.7588( 1−0.6

 −1
1 ,


 )3 } 3
1−0.3 3 1−0.7 3
1+{0.7588( ) +0.1771( ) +0.0428( 1−0.7 )3 +0.0214( 1−0.3


0.3 0.7 0.7 0.3
 −1
⟩, 1
1+{0.7588( 1−0.5
0.5
)3 +0.1771( 1−0.3
0.3
)3 +0.0428( 1−0.7
0.7
)3 +0.0214( 1−0.9
0.9
)3 } 3

BN DP W GA(b̃1 , b̃2 , b̃3 , b̃4 ) = ⟨0.4519, 0.6852, 0.5591, −0.7649, −0.3175, −0.4092⟩.

5.2 Bipolar neutrosophic Dombi prioritized ordered weighted


geometric aggregation operator
+ − − −
Definition 5.5. Let b˜i =< t+ +
i , µi , νi , ti , µi , νi > (i = 1, , 2, 3, ..., n) be a family of
BNNs. A mapping BNDPOWGA: U n → U is called bipolar neutrosophic Dombi priori-
tized ordered weighted geometric aggregation (BNDPOWGA) operator, if it satisfies

n ∧υ j ∧υ1 ∧υ2 ∧n
BN DP OW GA(b̃1 , b̃2 , ..., b̃n ) = b̃σ(j) = b̃σ(1)
D
⊗D b̃σ(2)
D
⊗D ... ⊗D b̃σ(n)
D
(13)
j=1

20
~j wj ∏j−1
where υj = , ~j = s(b̃i ) (i = 2, 3, ..., n), ~1 = 1 and s(b̃i ) is the value of
j=1 ~j wj
Σn i=1

score for b̃i . where σ is permutation that orders the elements: b̃σ(1) ≥ b̃σ(2) ≥ ... ≥ b̃σ(n) .
where w = (w1 , w2 , ..., wn )T is the weight vector of b̃i (i = 1, 2, ..., n), wi ∈ [0, 1] and
Σi=1 wi = 1.

+ − − −
Theorem 5.6. Let b˜i =< t+ +
i , µi , νi , ti , µi , νi > (i = 1, 2, 3, ..., n) be a family of BNNs
and w = (w1 , w2 , ..., wn )T be the weight vector of bi , wi ∈ [0, 1] and Σi=1 wi = 1. Then,
the aggregated value of BNDPOWGA operator is still a BNN, which is calculated by the
following formula
∧υ1 ∧υ2 ∧υ n
BNDPOWGA(b̃1 , b̃2 , ..., b̃n ) = b̃σ(1) D
⊗D b̃σ(2)D
⊗D ... ⊗D b̃σ(n) D



 ⟨ 1
,1 − 1
,1 − 1
,
 j wj
1−t+
 1+{Σnj=1 Σn ~ w ( + )l } l
~ j 1
1+{Σn
~ j w j
µ+
j 1
)l } l 1+{Σn
~j w j
ν+
j 1
)l } l
j=1 Σn ~ w ( + j=1 Σ n ~ w ( +
j=1 j j t j=1 j j 1−µj j=1 j j 1−νj
= j


1
− 1, −1
, −1


 1+{Σn ~j wj −t− 1
( j )l } l 1+{Σn
~j wj
(
1+µ−
j l 1
) } l 1+{Σn
~j wj
(
1+ν −
j l 1
) }l
j=1 Σn ~j wj
j=1 1+t− j=1 Σn ~j wj
j=1 −µ− j=1 Σn ~j wj
j=1 −ν −
j j j
(14)
By the mathematical induction, we can prove theorem (5.6).

Proof: If n = 2 based of the Dombi operations on BNNs in definition (5.5), we can


obtain the following result: ∧υ 1 ∧υ2
BNDPOWGA (b̃1 , b̃2 ) = b̃σ(1) D
⊗D b̃σ(2) D



 ⟨ 1
,1 − 1
,


~1 w1 1−t+1 )l + ~2 w2 1−t+ 1
2 )l } l ~1 w1 µ+
1 )l + ~2 w2 µ+
( 2 + )l } l
1


1+{ ~ w
1 1 +~ w
2 2
( +
t1 ~1 1w +~ w
2 2
( +
t2
1+{ ~ w
1 1 +~ w
2 2 1−µ
( + ~ w
1 1 +~ w
2 2 1−µ
 1 2

1− +
1
+ , −t−
1
−t−
− 1,
= ~1 w1 ν1 ~2 w2 ν2 1
l} l ~1 w1 ~2 w2 1
2 )l } l
 1+{ ( ) l+ ( ) 1+{ ( 1 )l + (


~1 w1 +~2 w2 1−ν + ~1 w1 +~2 w2 1−ν + ~1 w1 +~2 w2 1+t− ~1 w1 +~2 w2 1+t−

1 2 1 2
 −1 −1
 1+{ ~1 w1 ( 1+µ1 )l + ~2 w2 ( 1+µ2 )l } 1l 1+{ ~1 w1 ( 1+ν1 )l + ~2 w2 ( 1+ν2 )l } 1l ⟩,
 − − , − −
~1 w1 +~2 w2 −µ− ~1 w1 +~2 w2 −µ− ~1 w1 +~2 w2 −ν − ~1 w1 +~2 w2 −ν −
 1 2 1 2


 ⟨ 1
, 1 − 1
, 1 − 1
,
 +
 1+{Σ2j=1 Σ2~j w~j w ( 1−t+j )l } 1l 2
1+{Σj=1 2
~j wj
(
µ +
j
)}
l
1
l 2
1+{Σj=1 2
~j wj ν+
(Σ2j=1 j + )l } l
1
t Σ ~ w + Σ ~ w
j=1 j j j=1 j j 1−µj j=1 j j 1−ν
= j j


1
− 1, −1
, −1
⟩.

 1+{Σ2 ~j wj −t −
( j )l } l
1
1+{Σ2
~j wj
(
1+µ −
j l 1
) } l 1+{Σ2
~j wj
(
1+ν −
j l 1
) }l
j=1 Σ2 ~ w
j=1 j j 1+t− j=1 Σ2 ~ w
j=1 j j −µ− j=1 Σ2 ~ w
j=1 j j −ν −
j j j

If n = m, based on equation (14), then we have the following equation:


⊗ ∧υ j
BNDPOWGA(b̃1 , b̃2 , ..., b̃m ) = m b̃
 j=1 j
⟨

1
,1 − 1
,1 − 1
,
 1+{Σm ~j wj ( 1−t+j )l } 1l
 m ~j wj
1+{Σj=1 Σm ~ w (
µ+
j
)}
l
1
l
~j wj ν+ 1
1+{Σj=1 Σm ~ w ( j + )l } l
m
j=1 Σm ~j wj + +
j=1 t j=1 j j 1−µj j=1 j j 1−νj
= j
 −1 −1
 1+{Σm ~j wj ( −t−j )l } 1l − 1, 1+{Σm ~j wj ( 1+µ−j )l } 1l , 1+{Σm ~j wj ( 1+νj− )l } 1l ⟩.
1

 j=1 Σm ~j wj
j=1 1+t− j=1 Σm ~j wj
j=1 −µ− j=1 Σm ~j wj
j=1 −ν −
j j j

21
If n = m + 1, then there is following result:∧υ
⊗ j ⊗
∧υ
m+1
BNDPOWGA(b̃1 , b̃2 , ..., b̃m , b̃m+1 ) = m b̃ b̃
 j=1 σ(j) σ(m+1)

 ⟨ 1
,1 − 1
,1 − 1
,

 1+{Σm ~ j w j
(
1−t+
j
)
1
l} l 1+{Σ m ~ j w j
(
µ+
j
) l} l
1
1+{Σ m ~j w j
(
ν+
j
)l} l
1
m ~ w m ~ w m ~ w
j=1 Σ
j=1 j j t+ j=1 Σ
j=1 j j 1−µj
+ j=1 Σ
j=1 j j 1−νj
+
= j


1
− 1, 1
, 1


 1+{Σm ~j wj −t−
j 1 ~j wj 1+µ−
j l 1 ~j wj 1+ν −
j l 1
j=1 Σm ~j wj ( 1+t− ) } 1+{Σm j=1 Σm ~j wj ( −µ− ) } 1+{Σm j=1 Σm ~j wj ( −ν − ) }
l l l l
 j=1 j j=1 j j=1 j


 ⟨ 1
,1 − 1
,1 − 1
,
 1+{ ~m+1 wm+1 ( 1−t+m+1 )l } 1l
 ~
1+{ m+1
wm+1 µ+
( m+1 )l } l
1 ~
1+{ m+1
wm+1 ν+
( m+1
1
)l } l
Σ m+1 ~ w +
tm+1 Σ m+1 ~ w +
1−µm+1 Σ m+1 ~ w 1−νm+1+
⊗D j=1 j j j=1 j j j=1 j j



1
− 1, 1
, 1


 1+{ ~m+1 wm+1
(
−t−
j
)
1
l} l 1+{
~m+1 wm+1 1+µm+1 l 1
(

) } l 1+{
~m+1 wm+1 1+νm+1 l 1
(

) } l
− −
Σm+1 ~j wj 1+t−
m+1 ~ w −µm+1 m+1 ~ w −νm+1

Σ Σ
j=1 m+1 j=1 j j j=1 j j


 ⟨ 1
,1 − 1
,1 − 1
,

 1+{Σm+1 ~ j w j
1−t+
j 1
1+{Σm+1
~ j wj
µ+
j 1
1+{Σm+1
~j w j
ν+
j 1
j=1 Σm+1 ~ w ( t+ ) } j=1 Σm+1 ~ w ( 1−µ+ ) } j=1 Σm+1 ~ w ( 1−ν + ) }
l l l l l l
j=1 j j j j=1 j j j j=1 j j j
=


1
− 1, 1
, 1


 1+{Σm+1 ~j wj ( −tj )l } 1l

1+{Σm+1
~j wj
(
1+µ−
j l 1
) } l 1+{Σm+1
~j wj
(
1+ν −
j l 1
) }l
j=1 Σm+1 ~ w 1+t− j=1 Σm+1 ~ w −µ− j=1 Σm+1 ~ w −ν −
j=1 j j j j=1 j j j j=1 j j j
Hence, Theorem (5.6 is true for n = m + 1. Thus, equation (14) holds for all n.

The BNDPOWGA operator also contains the following properties:

+ − − −
(1) Idomopotency: Let all the BNNs be b˜i =< t+ +
i , µi , νi , ti , µi , νi >= b̃ (i = 1, 2, 3, ..., n),
where w = (w1 , w2 , ..., wn ) be the weight vector of b̃i , wi ∈ [0, 1] and Σi=1 wi = 1.
T

Then, BNDPOWGA(b̃1 , b̃2 , ..., b̃n ) = b̃.

(2) Monotonicity: Let b̃i (i = 1, 2, ..., n) and b̃,i (i = 1, 2, ..., n) be two families of BNNs,
where w = (w1 , w2 , ..., wn )T be the weight vector of b̃i and b̃,i , wi ∈ [0, 1] and Σi=1 wi =
1. For all i ,if b̃i ≥ b̃,i , then BNDPOWGA(b̃1 , b̃2 , ..., b̃n ) ≥ BNDPOWGA(b̃,1 , b̃,2 , ..., b̃,n ).
+ − − −
(3) Boundedness: Let b˜i =< t+ +
i , µi , νi , ti , µi , νi > (i = 1, , 2, 3, ..., n) be a family
of BNNs, where w = (w1 , w2 , ..., wn )T be the weight vector of bi , wi ∈ [0, 1] and
Σi=1 wi = 1. Therefore, we have
BNDPOWGA(b̃− , b̃− , ..., b̃− ) ≤ BNDPOWGA(b̃1 , b̃2 , ..., b̃n ) ≤ BNDPOWGA(b̃+ , b̃+ , ..., b̃+ ),
where b̃− =< t+ , µ+ , νb̃+− , t− , µ− , νb̃−− >
{ b̃− b̃− b̃− b̃−
< min(t+ + + + + + + +
1 , t2 , ...tn , ), max(µ1 , µ2 , ...µn ), max(ν1 , ν2 , ..., νn ),
+
=
max(t− − − − − − − −
1 , t2 , ...tn ), min(µi , µi , ...µi ), min(ν1 , ν2 , ..., νn ) >

and
b̃+ =< t+ , µ+ , νb̃++ , t− , µ− , νb̃−+ >
{ b̃+ b̃+ b̃+ b̃+
< max(t+ + + + + + + +
1 , t2 , ...tn , ), min(µ1 , µ2 , ...µn ), min(ν1 , ν2 , ..., νn ),
+
=
min(t− − − − − − − −
1 , t2 , ...tn ), max(µi , µi , ...µi ), max(ν1 , ν2 , ..., νn ) > .

22
+ − − −
Proof (1) Since b˜i =< t+ +
i , µi , νi , ti , µi , νi >= b̃(i = 1, , 2, 3, ..., n). Then, by using
equation (14), we can obtain the following result:

⊗ ∧υj
BNDPOWGA(b̃1 , b̃2 , ..., b̃n ) = nj=1 b̃j


 ⟨ 1
,1 − 1
,1 − 1
,
 n
1−t+
 1+{Σj=1 Σn ~ w ( + ) } l
~j wj j l
1
n ~j wj
1+{Σj=1 Σn ~ w (
µ+
j
)l }
1
l 1+{Σn ~j wj
(
ν+
j
)
1
l} l
+ j=1 Σ n ~ w +
j=1 j j t j=1 j j 1−µj j=1 j j 1−νj
= j


1
− 1, −1
, −1


 1+{Σn ~j wj −t−
( j )l } l
1
1+{Σn
~j wj
(
1+µ−
j l 1
) } l 1+{Σn
~j wj
(
1+ν −
j l 1
) }l
j=1 Σn ~j wj
j=1 1+t− j=1 Σn ~j wj
j=1 −µ− j=1 Σn ~j wj
j=1 −ν −
j j j



⟨ 1+{( 1−t+ )l } 1l , 1 −
1 1
µ+ 1 ,1 − 1
ν+
1 ,
1+{( )l } l 1+{( )l } l
= t+ 1−µ+ 1−ν +
=< t+ , µ+ , ν + , t− , µ− , ν − >= b̃.
 −1 −1
 1+{( −t− )l } 1l − 1, 1+{( 1+µ− )l } 1l , 1+{( 1+ν − )l } 1l ⟩
1

1+t− −µ− −ν −

Hence, BNDPOWGA(b̃1 , b̃2 , ..., b̃n ) = b̃ holds.


(2) based on the equation (14), the property is obvious.
(3) Let b̃− =< t+b̃−
, µ+
b̃−
, νb̃+− , t−
b̃−
, µ−
b̃−
, νb̃−− > and b̃+ =< t+
b̃+
, µ+
b̃+
, νb̃++ , t−
b̃+
, µ−
b̃+
, νb̃−+ >. There
are the following inequalities:
1
1−t+
≤ 1
~ w 1−t+
≤ 1
1−t+
,
~j wj 1 1
1+{Σn b̃− )l } l 1+{Σn j j j
)l } l ~j wj + l 1

j=1 Σn ~ w
(
1+
j=1 Σn ~ w
(
t+
1+{Σn
j=1 Σn ~ w
( ) }l
j=1 j j j=1 j j j=1 j j t+
b̃− j
b̃+

1− 1
µ+
≤ 1− ~j wj
1
µ+
≤ 1− 1
µ+
,
~j wj 1 1 ~j wj 1
1+{Σn
+
( b̃ + )l } l 1+{Σn ( j + )l } l 1+{Σn ( b̃− )l } l
j=1 Σn ~j wj j=1 Σn ~j wj j=1 Σn ~j wj
j=1 1−µ j=1 1−µ
j j=1 1−µ+
b̃+ −

1− 1
ν+
≤ 1− ~j wj
1
ν+
≤ 1− 1
ν+
,
~j wj 1 1 ~j wj 1
1+{Σn
+
( b̃ + )l } l 1+{Σn ( j + )l } l 1+{Σn ( b̃− )l } l
j=1 Σn ~j wj j=1 Σn ~j wj j=1 Σn ~j wj
j=1 1−ν j=1 1−ν
j j=1 1−ν +
b̃+ −

1
−t−
−1 ≤ ~j wj
1
−t− 1
−1 ≤ 1
−t−

~j wj 1 ~j wj − 1
1+{Σn
˜
+
( b− )l } l 1+{Σn
j=1 Σn ~j wj ( j− )l } l 1+{Σn
˜
( b− )l } l
j=1 Σn ~j wj 1+t j=1 1+t j=1 Σn ~j wj 1+t
j=1 j j=1
b̃+ −

1,
−1 −1 −1
1+µ−
≤ ~j wj 1+µ−
≤ 1+µ−
,
~j wj 1 j l 1 ~j wj 1
1+{Σn ( b̃− )l } l 1+{Σn ( ) }l 1+{Σn ( b̃+ )l } l
j=1 Σn ~j wj
j=1 −µ−
j=1 Σn ~j wj
j=1 −µ− j=1 Σn ~j wj
j=1 −µ−
b̃ − j b̃ +
−1 −1 −1
1+ν −
≤ ~j wj 1+ν −
≤ 1+ν −
.
~j wj 1 j l 1 ~j wj 1
1+{Σn ( b̃− )l } l 1+{Σn ( ) }l 1+{Σn ( b̃+ )l } l
j=1 Σn ~j wj
j=1 −ν −
j=1 Σn ~j wj
j=1 −ν − j=1 Σn ~j wj
j=1 −ν −
b̃ − j b̃ +

Hence, BNDPOWGA(b̃− , b̃− , ..., b̃− ) ≤ BNDPOWGA(b̃1 , b̃2 , ..., b̃n ) ≤ BNDPOWGA(b̃+ , b̃+ , ..., b̃+ )
holds.

Example 5.7. Let b˜1 =< 0.6, 0.7, 0.3, −0.6, −0.3, −0.5 >, b˜2 =< 0.5, 0.4, 0.6, −0.6, −0.7, −0.3 >,
b˜3 =< 0.6, 0.7, 0.4, −0.9, −0.7, −0.7 > and b˜4 =< 0.2, 0.6, 0.8, −0.6, −0.3, −0.9 > be
four BNNs and let the weight vector of BNNs b˜i (i = 1, 2, 3, 4) be w = ( 21 , 14 , 18 , 18 )T .
Now, by definition (11), ~1 = 1, ~2 = s(b̃1 ) = 0.4667, ~3 = s(b̃1 )s(b̃2 ) = 0.2256

23
and ~4 = s(b̃1 )s(b̃2 )s(b̃3 ) = 0.1128. w1 = 21 , w2 = 14 , w3 = 18 and w4 = 81 are the
weights of b̃i (i = 1, 2, 3, 4) such that Σi=1 wi = 1. By definition (11), Σ4~1 w~j1wj = 0.7588,
j=1
~2 w2 ~3 w3 ~4 w4
Σ4j=1 ~j wj
= 0.1771, Σ4j=1 ~j wj
= 0.0428, Σ4j=1 ~j wj
= 0.0214. Then, by theorem (5.6), for
l=3


 ⟨ 1
1 ,

 1+{0.7588( 1−0.6 3
) +0.1771( 1−0.5 3
) +0.0428( 1−0.6 )3 +0.0214( 1−0.2 )3 } 3


0.6 0.5 0.6 0.2

 1− 1
1 ,

 0.7 3 +0.1771( 0.4 )3 +0.0428( 0.7 )3 +0.0214( 0.6 )3 } 3


1+{0.7588( 1−0.7
) 1−0.4 1−0.7 1−0.6

1 − 1
1 ,
1+{0.7588( 1−0.4 ) +0.1771( 1−0.6 ) +0.0428( 1−0.3 ) +0.0214( 1−0.8 ) } 3
0.4 3 0.6 3 0.3 3 0.8 3
BN DP OW GA(b̃1 , b̃2 , b̃3 , b̃4 ) =


1
1 − 1,

 0.9 0.6
)3 +0.1771( 1+0.6 0.6
)3 +0.0428( 1+0.6 0.6
)3 +0.0214( 1+0.6 )3 } 3


1+{0.7588( 1+0.9

 −1
1 ,


 )3 } 3
1−0.7 3 1−0.7 3
1+{0.7588( ) +0.1771( ) +0.0428( 1−0.3 )3 +0.0214( 1−0.3


0.7 0.7 0.3 0.3
 −1
1 ⟩,
1+{0.7588( 1−0.7
0.7
)3 +0.1771( 1−0.3
0.3
)3 +0.0428( 1−0.5
0.5
)3 +0.0214( 1−0.9
0.9
)3 } 3

BN DP OW GA(b̃1 , b̃2 , b̃3 , b̃4 ) = ⟨0.4519, 0.6852, 0.5651, −0.8915, −0.5098, −0.4292⟩.

6 Model for MADM using Bipolar Neutrosophic fuzzy


information
In this section, three comprehensive MADM methods are put forward based on the pro-
posed BNDWGA, BNDOWGA, BNDPWGA and BNDPOWGA operators.
For MADM problems with bipolar neutrosophic information, let A = {A1 , A2 , ..., An }
be a set of alternatives and C = {C1 , C2 , ..., Cn } be a set of attributes or criteria. (For
BNDPWGA and BNDPOWGA operators, there is a prioritization between the attributes
expressed by the linear ordering C1 > C2 > ... > Cn , indicates attribute Cτ has a higher
priority than Cρ , if τ < ρ.) Let w = (w1 , w2 , ..., wn )T be the weight vector of attributes
such that wj > 0, Σj=1 wj = 1(j = 1, 2, ..., n) and wj refers to the weight of attribute Cj .
+ − − −
Suppose that B = (b̃ij )m×n = (t+ +
ij , µij , νij , tij , µij , νij )m×n (i = 1, 2, ..., m)(j = 1, 2, ..., n)
is BNN decision matrix, where t+ + +
ij , µij , νij indicates the truth membership degree, in-
determinacy membership degree and falsity membership degree of alternative Ai under
Cj with respect to positive preferences and t− − −
ij , µij , νij indicates the truth membership
degree, indeterminacy membership degree and falsity membership degree of alterna-
tive Ai under criteria Cj with respect to negative preferences. We have the conditions
+ − − − − − −
ij , µij , νij , tij , µij andνij ∈ [0, 1] such that 0 ≤ tij + µij + νij − tij − µij − νij ≤ 6 for
t+ + + + +

(i = 1, 2, ..., m) (j = 1, 2, ..., n).


An algorithm is designed to solve MADM problem using bipolar neutrosophic Dombi ag-
gregation operators are as BNDWGA, BNDOWGA, BNDPWGA and BNDPOWGA. We
can construct the algorithm as follows;
Algorithm

24
Step 1 Collect the bipolar neutrosophic evaluation information.
Step 2 Compute the score and accuracy values of the collected evaluation information.
The score values S(b̃ij ) and accuracy values A(b̃ij ) of alternatives Ai can be computed
based on Eqs. (1) and (2).
Step 3 Reorder evaluation information under each criteria.
The score values are utilized to rank b̃ij , such that S(b̃i(j) ) < S(b̃i(j+1) ). If S(b̃i(j) ) =
S(b̃i(j+1) ), then the accuracy values are utilized to rank b̃ij , such that A(b̃i(j) ) < A(b̃i(j+1) ).
Step 4 Derive the collective BNN b̃i (i = 1, 2, ..., m) for the alternative Ai (i = 1, 2, ..., m).
Method (1) Utilize BNDWGA operator to compute the collective BNN for each alterna-
tive, then
⊗ ∧wj
b̃i = BNDWGA(b̃i1 , b̃i2 , ..., b̃in ) = nj=1 b̃ij


⟨
1
,1 − 1
,1 − 1
,
 +
 1+{Σnj=1 wj ( 1−t+ij )l } 1l n
1+{Σj=1 wj (
µ+
ij
)}
l
1
l
ν+ 1
1+{Σj=1 wj ( ij+ )l } l
n
t 1−µ + 1−ν
= ij ij ij
(15)


1
− 1, −1
, −1
⟩,
 −
 1+{Σn w ( ij )l } l
−t 1
1+{Σn w (
1+µ −
ij l l
)}
1
1+{Σn w (
1+ν −
ij l l
)}
1
1+t− −µ− −ν −
j=1 j j=1 j j=1 j
ij ij ij

where w = (w1 , w2 , ..., wn ) is the weight vector such that wj ∈ [0, 1] and Σnj=1 wj = 1.
Method (2) Utilize BNDOWGA operator to compute the collective BNN for each alter-
native, then
⊗ ∧w
k
b̃i = BNDOWGA(b̃i1 , b̃i2 , ..., b̃in ) = nk=1 b̃σ(ik)


 ⟨ 1
,1 − 1
,1 − 1
,
 1+{Σnk=1 wk ( +ik )l } l
1−t+ 1
1+{Σn
k=1
µ+
wk ( ik+ )l } l
1
1+{Σn
k=1
ν+ 1
wk ( ik+ )l } l
t 1−µ 1−ν
= ik ik ik (16)


1
− 1, −1
, −1
⟩,
 1+{Σn wk ( ik )l } l
−t − 1
1+{Σn wk (
1+µ− 1
ik )l } l 1+{Σn wk (
1+ν − 1
ik )l } l
k=1 1+t− k=1 −µ− k=1 −ν −
ik ik ik

where σ is permutation that orders the elements: where σ is permutation that orders the
elements: b̃σ(i1) ≥ b̃σ(i2) ≥ ... ≥ b̃σ(in) . where w = (w1 , w2 , ..., wn )T is the weight vector
such that wk ∈ [0, 1] and Σnk=1 wk = 1.
Method (3) Calculate the values of ~ij (i = 1, 2, ..., m) (j = 1, 2, ..., n) as follows:


ij−1
~ij = s(b̃ip )(i = 1, 2, ..., m)(j = 2, 3, ..., n), (17)
ip=1

~i1 = 1(i = 1, 2, ..., m), (18)

and utilize BNDPWGA operator to compute the collective BNN for each alternative,
then ⊗ ∧υij
b˜i = BNDPWGA(b̃i1 , b̃i2 , ..., b̃in ) = nj=1 b̃ij

25


 ⟨ 1
,1 − 1
,1 − 1
,

 1+{Σnj=1 Σn~ij~wijw 1−t+ 1
( +ij )l } l 1+{Σn
~ij wij µ+ 1
( ij+ )l } l 1+{Σn
~ij wij
(
ν+
ij 1
)l } l
j=1 Σn ~ij wij j=1 Σn ~ij wij 1−ν +
j=1 ij ij t j=1 1−µ j=1
= ij ij ij


1
− 1, −1
, −1
⟩,

 1+{Σj=1 Σn ~ w
n ~ij wij −t− 1
( ij )l } l 1+{Σn
~ij wij
(
1+µ−
ij l 1
) }l 1+{Σn
~ij wij
(
1+ν −
ij l 1
) }l
j=1 ij ij 1+t− j=1 Σn ~ij wij
j=1 −µ− j=1 Σn ~ij wij
j=1 −ν −
ij ij ij
(19)
~ij wij T
where υij = and w = (w1 , w2 , ..., wn ) is the weight vector (i = 1, 2, ..., m) such
j=1 ~ij wij
Σn
that wj ∈ [0, 1] and Σj=1 wj = 1.

Method (4) Calculate the values of ~ik (i = 1, 2, ..., m) (k = 1, 2, ..., n) as follows:


ij−1
~ik = s(b̃ip )(i = 1, 2, ..., m)(j = 2, 3, ..., n), (20)
ip=1

~i1 = 1(i = 1, 2, ..., m), (21)


and utilize BNDPOWGA operator to compute the collective BNN for each alternative,
then ⊗ ∧υ
ik
b˜i = BNDPOWGA(b̃i1 , b̃i2 , ..., b̃in ) = nk=1 b̃ik


 ⟨ 1
,1 − 1
,1 − 1
,
 1+{Σnk=1 Σn~ik~wikw ( 1−t+ik )l } 1l
+ µ+ 1 ν+ 1
~ik wik ik )l } l ~ik wik
1+{Σn
k=1 Σ n ~ w
( + 1+{Σn
k=1 Σ n ~ w
( ik+ )l } l
k=1 ik ik t k=1 ik ik 1−µik k=1 ik ik 1−νik
= ik
 −1 −1
 1+{Σn n~ik wik ( −t−ik )l } 1l − 1, 1+{Σn n~ik wik ( 1+µ−ik )l } 1l , 1+{Σn n~ik wik ( 1+νik ⟩,
1
 − 1
)l } l
k=1 Σ ~ w
k=1 ik ik 1+t− k=1 Σ ~ w
k=1 ik ik −µ− k=1 Σ ~ w
k=1 ik ik −ν −
ik ik ik
(22)
~ik wik
where υik = and σ is permutation that orders the elements: b̃σ(i1) ≥ b̃σ(i2) ≥
k=1 ~ik wik
Σn
... ≥ b̃σ(in) . where w = (w1 , w2 , ..., wn )T is the weight vector such that wk ∈ [0, 1] and
Σk=1 wk = 1.
Step 5 Calculate the score S(b̃i ) (i = 1, 2, ..., m) of BNNs b̃i (i = 1, 2, ..., m) to rank all the
alternatives Ai (i = 1, 2, ..., m) and then to select the best one(s). If there is no difference
between two scores S(b̃i ) and S(b̃j ). Then, we need to calculate the accuracy degrees
A(b̃i ) and A(b̃j ) of overall BNNs b̃i and b̃j , respectively, and then rank the alternatives Ai
and Aj in accordance with accuracy degrees A(b̃i ) and A(b̃j ).
Step 6 Rank all the alternatives Ai (i = 1, 2, ..., m) and select the best one(s) in accordance
with S(b̃i ) (i = 1, 2, ..., m).
Step 7 End.

7 Numerical Example
A numerical example of the selection of cultivating crop taken from Deli et al. [10] is
provided here. Furthermore, parametric analysis and comparative analysis confirm the

26
flexibility and effectiveness of the proposed methods. In ordered to increase the produc-
tion of agriculture, an agriculture department considers the selecting a crop for cultivating
in the farm. Four cultivating crops A1 , A2 , A3 , and A4 are selected for further evalua-
tion through preliminary screening. The agriculture department invited four groups of
experts to evaluate the information. The experts include investment experts, land ex-
perts, weather experts and human resource professionals. The four cultivating crops are
evaluated by experts on the basis of four attributes or criteria: cost (C1 ), nutrition of
land (C2 ), effect of weather (C3 ), labor (C4 ). (For the proposed methods BNDPWGA
and BNDPOWGA operators, the prioritization relation for the attributes or criteria is
given as: C1 ≻ C2 ≻ C3 ≻ C4 .) These attributes or criteria are interactive and interlinked.

Algorithm
Step 1 Collect the bipolar neutrosophic evaluation information.
The evaluation information obtained through expert discussion is shown in Table 1.

Table 1 Bipolar neutrosophic evaluation information


C1 C2 C3 C4
A1 ⟨0.5, 0.7, 0.2, −0.7, −0.3, −0.6⟩ ⟨0.4, 0.4, 0.5, −0.7, −0.8, −0.4⟩ ⟨0.7, 0.7, 0.5, −0.8, −0.7, −0.6⟩ ⟨0.1, 0.5, 0.7, −0.5, −0.2, −0.8⟩
A2 ⟨0.9, 0.7, 0.5, −0.7, −0.7, −0.1⟩ ⟨0.7, 0.6, 0.8, −0.7, −0.5, −0.1⟩ ⟨0.9, 0.4, 0.6, −0.1, −0.7, −0.5⟩ ⟨0.5, 0.2, 0.7, −0.5, −0.1, −0.9⟩
A3 ⟨0.3, 0.4, 0.2, −0.6, −0.3, −0.7⟩ ⟨0.2, 0.2, 0.2, −0.4, −0.7, −0.4⟩ ⟨0.9, 0.5, 0.5, −0.6, −0.5, −0.2⟩ ⟨0.7, 0.5, 0.3, −0.4, −0.2, −0.2⟩
A4 ⟨0.9, 0.7, 0.2, −0.8, −0.6, −0.1⟩ ⟨0.3, 0.5, 0.2, −0.5, −0.5, −0.2⟩ ⟨0.5, 0.4, 0.5, −0.1, −0.7, −0.2⟩ ⟨0.4, 0.2, 0.8, −0.5, −0.5, −0.6⟩

Step 2 Obtain the score values and accuracy values of the collected evaluation infor-
mation.
The score values S(b̃ij ) and accuracy values A(b̃ij ) of alternative Ai under criterion Cj
can be obtained based on Eqs. (1) and (2). S(b̃ij ) are shown in Table 2 and 3, respectively.

C1 C2 C3 C4
A1 0.4667 0.5000 0.5000 0.4000
Table 2 Score values S(b̃ij ) A2 0.4667 0.3667 0.6667 0.5167
A3 0.5167 0.5833 0.5000 0.4833
A4 0.4833 0.4667 0.5667 0.5000

C1 C2 C3 C4
A1 0.2000 −0.4000 0 −0.3000
Table 3 Accuracy values A(b̃ij ) A2 −0.2000 −0.7000 0.7000 0.2000
A3 0.2000 0 0 0.2000
A4 0 −0.2000 0.1000 −0.3000

Step 3 Reorder evaluation information under each criteria.


The score values are utilized to rank b̃ij , such that S(b̃i(j) ) < S(b̃i(j+1) ). If S(b̃i(j) ) =
S(b̃i(j+1) ), then the accuracy values are utilized to rank b̃ij , such that A(b̃i(j) ) < A(b̃i(j+1) ).
Subsequently, b̃ij is reordered on the basis of S(b̃ij ) and A(b̃ij ) (Table 4).

27
Table 4 Bipolar neutrosophic evaluation information
C1 C2 C3 C4
A1 ⟨0.7, 0.7, 0.5, −0.8, −0.7, −0.6⟩ ⟨0.4, 0.4, 0.5, −0.7, −0.8, −0.4⟩ ⟨0.5, 0.7, 0.2, −0.7, −0.3, −0.6⟩ ⟨0.1, 0.5, 0.7, −0.5, −0.2, −0.8⟩
A2 ⟨0.9, 0.4, 0.6, −0.1, −0.7, −0.5⟩ ⟨0.5, 0.2, 0.7, −0.5, −0.1, −0.9⟩ ⟨0.9, 0.7, 0.5, −0.7, −0.7, −0.1⟩ ⟨0.7, 0.6, 0.8, −0.7, −0.5, −0.1⟩
A3 ⟨0.2, 0.2, 0.2, −0.4, −0.7, −0.4⟩ ⟨0.3, 0.4, 0.2, −0.6, −0.3, −0.7⟩ ⟨0.9, 0.5, 0.5, −0.6, −0.5, −0.2⟩ ⟨0.7, 0.5, 0.3, −0.4, −0.2, −0.2⟩
A4 ⟨0.5, 0.4, 0.5, −0.1, −0.7, −0.2⟩ ⟨0.4, 0.2, 0.8, −0.5, −0.5, −0.6⟩ ⟨0.9, 0.7, 0.2, −0.8, −0.6, −0.1⟩ ⟨0.3, 0.5, 0.2, −0.5, −0.5, −0.2⟩

Step 4 Derive the collective BNN b̃i (i = 1, 2, ..., m) for the alternative Ai (i = 1, 2, ..., m).
Method (1) Utilize BNDWGA operator using Eq. (15) and supporting l = 7 to compute
the collective BNN for each alternative, then

b̃1 = ⟨0.1301, 0.6857, 0.6345, −0.7517, −0.2494, −0.4480⟩;


b̃2 = ⟨0.5735, 0.6795, 0.7667, −0.6913, −0.1301, −0.1038⟩;
b̃3 = ⟨0.2324, 0.4546, 0.4264, −0.5839, −0.2494, −0.2335⟩;
b̃4 = ⟨0.3424, 0.6788, 0.7482, −0.7837, −0.5323, −0.1092⟩.
Method (2) Utilize BNDOWGA operator using Eq. (16) and supporting l = 7 to compute
the collective BNN for each alternative, then

b̃1 = ⟨0.1301, 0.6857, 0.6345, −0.7841, −0.2512, −0.4480⟩;


b̃2 = ⟨0.5493, 0.6357, 0.7496, −0.6569, −0.1193, −0.1193⟩;
b̃3 = ⟨0.2160, 0.4527, 0.4264, −0.5661, −0.2506, −0.2335⟩;
b̃4 = ⟨0.3626, 0.6343, 0.7664, −0.7482, −0.5342, −0.1298⟩.

Method (3) Calculate the values of ~ij (i = 1, 2, ..., m) (j = 1, 2, ..., n) using Eqs. (17)
and (18) as follows:

 
1.0000 0.4667 0.2333 0.1167
1.0000 0.4667 0.1711 0.1141
~ij = 
1.0000

0.5167 0.3014 0.1507
1.0000 0.4833 0.2256 0.1278
and utilize BNDPWGA operator using Eq. (19) and supporting l = 7 to compute the
collective BNN for each alternative, then

b̃1 = ⟨0.1608, 0.6933, 0.5760, −0.7305, −0.2840, −0.4601⟩;

b̃2 = ⟨0.6325, 0.6923, 0.7578, −0.6983, −0.1610, −0.1007⟩;


b̃3 = ⟨0.2387, 0.4263, 0.3979, −0.5917, −0.2807, −0.2631⟩;
b̃4 = ⟨0.3533, 0.6914, 0.7013, −0.7934, −0.5494, −0.1037⟩.

28
Method (4) Calculate the values of ~ik (i = 1, 2, ..., m) (j = 1, 2, ..., n) using Eqs. (20)
and (21) as follows:

 
1.0000 0.4667 0.2333 0.1167
1.0000 0.4667 0.1711 0.1141
~ik = 
1.0000

0.5167 0.3014 0.1507
1.0000 0.4833 0.2256 0.1278
and utilize BNDPOWGA operator using Eq. (22) and supporting l = 7 to compute the
collective BNN for each alternative, then
b̃1 = ⟨0.1608, 0.6933, 0.5788, −0.7937, −0.2999, −0.4601⟩;
b̃2 = ⟨0.5612, 0.5900, 0.7046, −0.6067, −0.1244, −0.1441⟩;
b̃3 = ⟨0.4097, 0.5978, 0.7582, −0.7181, −0.5555, −0.1441⟩;
b̃4 = ⟨0.2072, 0.4162, 0.3979, −0.5511, −0.2903, −0.2629⟩.

Step 5 Calculate the score values S(b̃i ) (i = 1, 2, ..., m) of BNNs b̃i (i = 1, 2, ..., m) for
each alternatives Ai (i = 1, 2, ..., m). Equation (1) is used to calculate the score value
S(b̃i ).

M ethod 1 The score values obtained using the BNDWGA operator are as follows.
S(b̃1 ) = 0.2926; S(b̃2 ) = 0.2783; S(b̃3 ) = 0.3751; S(b̃4 ) = 0.2955.

M ethod 2 The score values obtained using the BNDOWGA operator are as follows.
S(b̃1 ) = 0.2875; S(b̃2 ) = 0.2910; S(b̃3 ) = 0.3758; S(b̃4 ) = 0.3129.

M ethod 3 The score values obtained using the BNDPWGA operator are as follows.
S(b̃1 ) = 0.3175; S(b̃2 ) = 0.2910; S(b̃3 ) = 0.3944; S(b̃4 ) = 0.3034.

M ethod 4 The score values obtained using the BNDPOWGA operator are as follows.
S(b̃1 ) = 0.3091; S(b̃2 ) = 0.3214; S(b̃3 ) = 0.3992; S(b̃4 ) = 0.3398.

Step 6 Rank all the alternatives Ai (i = 1, 2, ..., m) and select the best one(s).
The alternative can be ranked in descending order on the basis of score values and the
comparison method in definition 4, and the best alternative can be selected.
M ethod 1 According to score values obtained using the BNDWGA operator, the ranking
order is A3 ≻ A4 ≻ A1 ≻ A2 . Thus, A3 is best.
M ethod 2 According to score values obtained using the BNDOWGA operator, the rank-
ing order is A3 ≻ A4 ≻ A2 ≻ A1 . Thus, A3 is best.

29
M ethod 3 According to score values obtained using the BNDPWGA operator, the ranking
order is A3 ≻ A1 ≻ A4 ≻ A2 . Thus, A3 is best.
M ethod 4 According to score values obtained using the BNDPOWGA operator, the rank-
ing order is A3 ≻ A4 ≻ A2 ≻ A1 . Thus, A3 is best.
Step 7 End.

8 Parametric analysis and Comparative analysis


This section describes the effect of parametric l on decision making results and comparison
between proposed methods and existing methods.

8.1 Analysis on the effect of parameter l on decision making


results
This subsection discusses the effect of parameter l in detail.
First, the effect of parameters l on the proposed operators is discussed. Table 5 Rankings
with parameter of the BNDWGA, BNDOWGA, BNDPWGA, and BNDPOWGA opera-
tors
l BN DW GA BN DOW GA BN DP W GA BN DP OW GA
1 A3 ≻ A4 ≻ A1 ≻ A2 A3 ≻ A4 ≻ A2 ≻ A1 A3 ≻ A1 ≻ A4 ≻ A2 A2 ≻ A3 ≻ A4 ≻ A1
2 A3 ≻ A1 ≻ A4 ≻ A2 A3 ≻ A4 ≻ A2 ≻ A1 A3 ≻ A1 ≻ A4 ≻ A2 A3 ≻ A2 ≻ A4 ≻ A1
3 A3 ≻ A1 ≻ A4 ≻ A2 A3 ≻ A4 ≻ A2 ≻ A1 A3 ≻ A1 ≻ A4 ≻ A2 A3 ≻ A4 ≻ A2 ≻ A1
4 A3 ≻ A1 ≻ A4 ≻ A2 A3 ≻ A4 ≻ A2 ≻ A1 A3 ≻ A1 ≻ A4 ≻ A2 A3 ≻ A4 ≻ A2 ≻ A1
5 A3 ≻ A1 ≻ A4 ≻ A2 A3 ≻ A4 ≻ A2 ≻ A1 A3 ≻ A1 ≻ A4 ≻ A2 A3 ≻ A4 ≻ A2 ≻ A1
6 A3 ≻ A4 ≻ A1 ≻ A2 A3 ≻ A4 ≻ A2 ≻ A1 A3 ≻ A1 ≻ A4 ≻ A2 A3 ≻ A4 ≻ A2 ≻ A1
7 A3 ≻ A4 ≻ A1 ≻ A2 A3 ≻ A4 ≻ A2 ≻ A1 A3 ≻ A1 ≻ A4 ≻ A2 A3 ≻ A4 ≻ A2 ≻ A1
8 A3 ≻ A4 ≻ A1 ≻ A2 A3 ≻ A4 ≻ A2 ≻ A1 A3 ≻ A1 ≻ A4 ≻ A2 A3 ≻ A4 ≻ A2 ≻ A1
9 A3 ≻ A4 ≻ A1 ≻ A2 A3 ≻ A4 ≻ A2 ≻ A1 A3 ≻ A1 ≻ A4 ≻ A2 A3 ≻ A4 ≻ A2 ≻ A1
10 A3 ≻ A4 ≻ A1 ≻ A2 A3 ≻ A4 ≻ A2 ≻ A1 A3 ≻ A1 ≻ A4 ≻ A2 A3 ≻ A4 ≻ A2 ≻ A1
Table 5 presents the corresponding ranking with respect to the BNDWGA, BNDOWGA,
BNDPWGA, and BNDPOWGA operators as the value of l changes from 0 to 10.

From Table 5, it is noticeable that when the value of l is changed for BNDWGA operator,
the ranking order is slightly different, but corresponding best alternative is identical. For,
1 ≤ l ≤ 10, the ranking orders are shown as A3 ≻ A4 ≻ A1 ≻ A2 and A3 ≻ A1 ≻ A4 ≻ A2 .
It follows that the best alternative is A3 .

From Table 5, for BNDOWGA operator shows when the value of l is changed, ranking
order is stable and the corresponding best alternative is identical. For, 1 ≤ l ≤ 10, then
the corresponding ranking order A3 ≻ A4 ≻ A2 ≻ A1 , the best alternative is A3 . Thus,

30
the best stable alternative is A3 .
From Table 5, for BNDPWGA operator when the value of l is changed, ranking order
is stable and the corresponding best alternative is identical. For, 1 ≤ l ≤ 10 then cor-
responding ranking orders A3 ≻ A1 ≻ A4 ≻ A2 , then best one is A3 . Thus, best stable
alternative is A3 .
From Table 5, for BNDPOWGA operator when the value of l is changed, the ranking
order is different. For, 1 ≤ l ≤ 2, the ranking orders are A2 ≻ A3 ≻ A4 ≻ A1 and
A3 ≻ A2 ≻ A4 ≻ A1 . It follows that the best alternatives are A2 and A3 , respectively.
For, 3 ≤ l ≤ 10, the ranking order is A3 ≻ A4 ≻ A2 ≻ A1 , then the best one is A3 .

8.2 comparative analysis


This subsection provides a comparative analysis of existing methods under the bipolar
neutrosophic set to verify the effectiveness and advantages of the proposed methods using
the same evaluation information as shown in the Table 1.
The proposed MADM methods are based on the proposed operators, a comparison of the
existing MADM methods is discussed based on the bipolar neutrosophic Dombi aggrega-
tion operators.
Table 6 Ranking obtained by different methods
Methods Rankings
AW operator [11] A3 ≻ A4 ≻ A2 ≻ A1
GW operator [11] A3 ≻ A4 ≻ A2 ≻ A1
BN-TOPSIS method(W1 ) [12] A4 ≻ A2 ≻ A3 ≻ A1
BN-TOPSIS method(W2 ) [12] A3 ≻ A2 ≻ A4 ≻ A1
FBNCWBM operator (s, t = 1) [9] A3 ≻ A4 ≻ A2 ≻ A1
FBNCGBM operator (s, t = 1) [9] A3 ≻ A4 ≻ A2 ≻ A1
BNDWGA operator (l = 7) A3 ≻ A4 ≻ A1 ≻ A2
BNDOWGA operator (l = 7) A3 ≻ A4 ≻ A2 ≻ A1
BNDPWGA operator (l = 7) A3 ≻ A1 ≻ A4 ≻ A2
BNDPOWGA operator (l = 7) A3 ≻ A4 ≻ A2 ≻ A1
The rankings obtained by the proposed methods show that A3 is best alternative.

Table 7 Characteristic comparison of different methods

31
Methods Flexible measure easier prioritized
AW operator [11] No No
GW operator [11] No No
BN-TOPSIS method(W1 ) [12] No No
BN-TOPSIS method(W2 ) [12] No No
FBNCWBM operator (s, t = 1) [9] Yes No
FBNCGBM operator (s, t = 1) [9] Yes No
BNDWGA operator (l = 7) Yes No
BNDOWGA operator (l = 7) Yes No
BNDPWGA operator (l = 7) Yes Yes
BNDPOWGA operator (l = 7) Yes Yes

In similar to Aw , Gw , BN-TOPSIS(W1 ) and BN-TOPSIS(W2 ) methods, the proposed


method based on proposed BNDWGA operator does not consider the interaction and in-
terrelation among attributes. In contrast to Aw , Gw , BN-TOPSIS(W1 ) and BN-TOPSIS(W2 )
method, the proposed BNDWGA operator can select the appropriate parameters accord-
ing to the preferences of the DMs. In similar to FBNCWBM (s, t = 1) and FBNCGBM
(s, t = 1) operators, the proposed BNDWGA operator can select the appropriate param-
eters according to preferences of the DMs. In contrast to FBNCWBM (s, t = 1) and
FBNCGBM (s, t = 1) operators, the proposed BNDWGA operator does not consider
the interaction and interrelation among attributes. Some MADM problems are indepen-
dent of attributes. Then, DMs can use the proposed method based on the proposed
BNDWGA operator to solve the MADM problems which are independent of attributes.
In contrast to Aw , Gw , BN-TOPSIS(W1 ) and BN-TOPSIS(W2 ) methods, the proposed
method based on proposed BNDOWGA operator considers the interaction and interrela-
tionship among the attributes by using ordered weighted geometric aggregation operator
and can select the appropriate parameters according to preferences of DMs. In similar to
FBNCWBM (s, t = 1) and FBNCGBM (s, t = 1) operators, the proposed BNDOWGA
operator can select the appropriate the parameters according to preferences of DMs and
can consider the interaction and interrelationship among the attributes. Some MADM
problems are dependent of attributes. Then, DMs can use the proposed method based
on the proposed BNDOWGA operator to solve these MADM problem which are de-
pendent of attributes. In contrast to Aw , Gw , BN-TOPSIS(W1 ) and BN-TOPSIS(W2 ),
FBNCWBM (s, t = 1) and FBNCGBM (s, t = 1) methods, the proposed method based
on proposed BNDPWGA operator considers the prioritized relationship among the at-
tributes by establishing the prioritized aggregation operator. In some practical MADM
problems, the prioritization relationship exists among attributes. Then, DMs can use
the proposed method based on proposed BNDPWGA operator to solve MADM problems
which have prioritization relationship among attributes. In contrast to Aw , Gw , BN-
TOPSIS(W1 ) and BN-TOPSIS(W2 ), FBNCWBM (s, t = 1) and FBNCGBM (s, t = 1)
methods, the proposed method based on proposed BNDPOWGA operator considers the
prioritized relationship among the attributes by establishing the prioritized aggregation

32
operator, and the interaction and interrelationship among attributes by using ordered
weighted geometric aggregation operator. In some practical MADM problem, the priori-
tization relationship, interaction and interrelationship exist among the attributes. Then,
DMs can use the proposed method based on the proposed BNDPOWGA operator to solve
the MADM problems which have the prioritization relationship, interaction and interre-
lationship among the attributes.
Thus, the proposed methods based on the proposed operators are more reliable and flexi-
ble in MADM problems. The DMs can use these proposed methods based on the proposed
operators according to their requirements in practical MADM problems.

9 conclusion
BNSs can describe fuzzy, bipolar, uncertain, and inconsistent information. The BND-
WGA, BNDOWGA, BNDPWGA, and BNDPOWGA operators were defined based on
Dombi operations to ensure that the bipolar neutrosophic aggregation operator is reliable
and flexible. Furthermore, a numerical example was given to prove the proposed methods.
Finally, the flexibility and reliability of the proposed methods were further demonstrated
through a parameter analysis and a comparative analysis with other methods.
The contribution of this paper are as follows. First, BNSs were used to present the deci-
sion making evaluation information. Second, Dombi operations were extended to bipolar
neutrosophic environment. Third, the BNDWGA, BNDOWGA, BNDPWGA and BND-
POWGA operators are proposed under the bipolar neutrosophic environment. Fourth,
MADM methods were developed based on the proposed operators. Finnaly, the flexibility,
practicality and efficiency of proposed methods were confirmed by a numerical example
and a comparative analysis.
In future, the BNDWGA, BNDOWGA, BNDPWGA and BNDPOWGA operators can be
extended to other fuzzy environments such as bipolar neutrosophic soft expert sets and
bipolar interval neutrosophic sets.

References
[1] Zadeh, L.A.: Fuzzy set,Inf. Control 8(3), 338-353 (1965)

[2] Atanassov, K.T.;: Ituitionistc fuzzy sets. Fuzzy Sets syst. 20(1), 87-96 (1986)

[3] Atanassov, K., Gargov, G.: Interval valued intuitionistic fuzzy sets. Fuzzy Sets Syst.
31(3), 343-349 (1989)

[4] Smarandache, F.: A unifying field in logics: neutrosophic logic. Multiple-valued Logic
8(3), 489-503 (1999)

33
[5] Wang, H., Madiraju, P., Zhang, Y., Sunderrraman, R.: Interval neutrosophic sets.
Mathematics 1, 274-277 (2004)

[6] Wang, H., Samarandache, F., Zhang, Y., Sunderraman, R.: Single valued neutro-
sophic sets. Rev. Air Force Acad. 17, 10-14 (2010)

[7] Gruber, J., seigel, E.H., Purcell, A.L., Earls, H.A., Cooper, G., Barrett, L.F.: Unseen
positive and negative affective information influences social perception in bipolar I
disorder and healthy adults. J. Affect. Disord. 192, 191-198 (2015)

[8] Bistarelli, S., Pini, M.S., F., Venable, K.B.: Modelling and solving bipolar preference
problems. Ercim Workshop on constraints Lisbona Giugno 23(4), 545-575 (2015)

[9] Wang, L., Zhang, H.-Y., Wang, J.-Q.: Frank Choquet Bonferroni Mean Operators of
Bipolar Neutrosophic Sets and Their Application to Multi-criteria Decision-making
Problems. In: Fuzzy Syst. 20(1): 13-28 (2018)

[10] Zhang, W.R.: Bipolar fuzzy setss and relations: a computational framework for
cognitive modeling and multiagent decision analysis. In: The Workshop on Fuzzy
Information Processing Society Bi conference, pp. 305-309 (1994)

[11] Deli, I., Ali, M., Smarandache, F.: Bipolar neutrosophic sets and their application
based onn multi-creteria decision making problems. International Conference on Ad-
vanced Mechatronic Systems, pp. 249-254 (2015)

[12] Dey, p.p., Pramanik, S., Giri, B.C.: TOPSIS for solving multi-attribute decision mak-
ing problems under bipolar neutrosophic environment, Pons asbl, Brussels, European
Union, pp. 65-77 (2016)

[13] Dombi, J.A general class of fuzzy operators, the demorgan class of fuzzy operators
and fuzziness measures induced by fuzzy operators. Fuzzy Sets Syst. 8, pp. 149-163
(1982)

[14] Yager, R. R. Prioritized aggregation operators. International Journal of Approximate


Reasoning, 48, pp. 263-274 (2008)

[15]

[16]

[17]

[18]

[19]

[20]

34
[21]

[22]

[23]

[24]

[25]

35

You might also like