Received September 4, 2019, accepted September 15, 2019, date of publication September 20, 2019,

date of current version October 3, 2019.

Digital Object Identifier 10.1109/ACCESS.2019.2942766

New Similarity Measures of Pythagorean

Fuzzy Sets and Their Applications
QIANG ZHANG 1, JUNHUA HU1 , JINFU FENG1 , AN LIU1 , AND YONGLI LI2,3
1 AeronauticsEngineering College, Air Force Engineering University, Xi’an 710038, China
2 Key Laboratory of Marine Intelligent Equipment and System of Ministry of Education, Shanghai Jiao Tong University, Shanghai 200240, China
3 Equipment Engineering College, Engineering University of CAPF, Xi’an 710086, China

Corresponding author: Qiang Zhang (zjslwdyx@163.com)

This work was supported in part by the National Natural Science Foundation of China under Grant 51779263.

ABSTRACT Similarity measure, as a tool to measure the similarity degree between two objects, is an
important research content in fuzzy set theory. Pythagorean fuzzy set, as a new extension of fuzzy set
theory, has been widely used in various fields. It is very necessary to study the similarity measure of the
Pythagorean Fuzzy set. Considering that the existing similarity measures cannot distinguish the highly
similar but inconsistent Pythagorean fuzzy sets and the calculation results are error-prone in application, this
paper introduces the exponential function to propose several new similarity measures of the Pythagorean
fuzzy set. Firstly, on the premise of introducing the existing similarity measures, several new similarity
measures are defined and their properties are discussed, and then the weighted similarity measures are
defined. Then, the new similarity measures and the existing similarity measures are compared by an example,
and it is verified that the new similarity measures can effectively distinguish highly similar but inconsistent
Pythagorean fuzzy sets. Finally, through three simulation cases, it is verified that the new similarity measures
can deal with different practical application problems more accurately and reliable than the existing similarity
measures.

INDEX TERMS Similarity measure, pythagorean fuzzy set, pythagorean fuzzy number, intuitionistic fuzzy
set, ranking method.

I. INTRODUCTION Yang [5] proposed two operations of the Pythagorean fuzzy

As an extension of the fuzzy set (FS), the intuitionistic
fuzzy set (IFS) has been widely studied after Atanassov [1]
proposed it. The IFS uses membership function and non-
membership function to characterize fuzzy set, and the sum of
membership degree (MD) and non-membership degree (ND)
is less than or equal to 1. On the basis of the IFS, Yager and
Abbasov [2], [3] relaxed the relationship between the MD and
the ND, and specified that the square sum of the MD and the
ND is less than or equal to 1, thus proposed the Pythagorean
fuzzy set (PFS).
The PFS expands the application range of the IFS and
attracts great attention of researchers. Ejegwa [4] explored the
concept of the PFS and deduced some theorems in connection
to the score function and the accuracy function, and then
presented a decision making (DM) approach of the career
placements by using the Pythagorean fuzzy relation. Peng and
The associate editor coordinating the review of this manuscript and
approving it for publication was Alba Amato.
numbers (PFNs) and discussed some desirable properties of
the Pythagorean fuzzy aggregation operators, and then initi-
ated a Pythagorean fuzzy superiority and inferiority ranking
method to deal with Internet stock investment problem. Peng
and Yuan [6] proposed several point operators for the PFNs
which can reduce the indeterminacy degree of the PFNs,
and introduced some generalized Pythagorean fuzzy aggrega-
tion operators. Garg [7] developed the correlation coefficient
based on the PFNs to solve the multiple criteria decision
making (MCDM) problem. Yager [8] proposed four kinds of
aggregation operators of the PFSs and applied them to handle
MCDM problems. Peng and Selvachandran [9] provided two
novel algorithms in DM problems under the Pythagorean
fuzzy environment. Peng [10] initiated some new operators
for the PFS and discussed their properties in detail, and
then presented an algorithm for solving the MCDM prob-
lem based on the weighted distance-based approximation. In
the research process of the PFS, some improvements have
been made. Peng and Li [11] presented two interval-valued

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see http://creativecommons.org/licenses/by/4.0/
138192 VOLUME 7, 2019
Q. Zhang et al.: New Similarity Measures of Pythagorean FSs and Their Applications
Pythagorean fuzzy DM methods and applied them in emer-
gency DM issues. Lu et al. [12] utilized hamacher opera-
tions to develop some hesitant Pythagorean fuzzy aggregation
operators, and then used these operators to develop some
methods to solve the hesitant Pythagorean fuzzy MCDM
problems. Grag [13] presented a new linguistic PFS by com-
bining the concepts of the PFS and the linguistic fuzzy set.
Zhang and Li [14] developed the Pythagorean fuzzy rough
set by combining the classical rough set and the Pythagorean
fuzzy relation.
The similarity measure is an important research content
in the FS theory and can be used to determine the similar-
ity degree between two objects. The similarity measures of
the IFS and the PFS are widely used in many fields, such
as the pattern recognition [15]–[21], the medical diagno-
sis [22]–[27] and the DM [28]–[34].
At present, there are many studies on the similarity mea-
sures of the IFS. Hung and Yang [35] reviewed several popu-
lar similarity measures between the FS and then extend those
similarity measures to the IFS, and then proposed two new
similarity measures between the IFSs. Li and Cheng [20]
proposed several new similarity measures of the IFS and
applied them to the pattern recognition. Liu [36] considered
Li and Cheng’s similarity measures are not reasonable in
some cases, thus presented several new modified similarity
measures between the IFSs. Szmidt and Kacprzyk [37] pro-
posed a new similarity measure of the IFS, and used it to
analyze the agreement extent in a group of experts. Hung and
Yang [38] presented a new method to calculate the similarity
measures between the IFSs based on the Hausdorff distance
concept. Hung and Yang [39] proposed several reasonable
measures induced by Lp metric to calculate the similarity
measure between the IFSs. Xia and Xu [40] proposed a series
of similarity measures of the IFS based on the intuitionistic
fuzzy operators. Xu [33] developed some similarity measures
of the IFS and applied them to the MCDM under intuition-
istic fuzzy environment. Ye [41] proposed two new cosine
similarity measures and weighted cosine similarity measures
between the IFSs based on the cosine function and the infor-
mation carried by the MD, the ND and the hesitation degree
(HD) in the IFS. Tian [26] presented a new fuzzy cotangent
similarity measure of the IFS, and then applied it to solve the
medical diagnosis problem. Muthukumar and Krishnan [22]
proposed a new similarity measure and a weighted similarity
measure of the intuitionistic fuzzy soft set and discussed
their basic properties. Hwang et al. [42] proposed a new
similarity measure of the IFS induced by the Jaccard index,
and applied it to the clustering problem. Based on the centroid
points of transformed right-angled triangular fuzzy numbers,
Chen et al. [43] proposed a new similarity measure between
the intuitionistic fuzzy values.
It can be seen from the above that the similarity measures
of the IFS have been successfully applied to different fields,
but there are some cases in practical application which cannot
be solved by the IFS. For example, in DM problems, if the
decision maker gives the MD and the ND as 0.7 and 0.5,
VOLUME 7, 2019
respectively, the IFS cannot be used. The PFS is an extension
of the IFS, which can deal with some cases where the IFS
cannot be used. Therefore, the application of the similarity
measures of the PFS will be more extensive. The study of the
similarity measures of the IFS can provide ideas and methods
for the study of the similarity measures of the PFS.
The studies on the similarity measures of the PFS are
relatively rare. Ejegwa [44] introduced the axiomatic defi-
nition of the similarity measure of the PFS. Zeng et al. [45]
presented some similarity measures of the PFS based on
several distance measures. Peng et al. [21] constructed the
definitions of the Pythagorean fuzzy information measures,
and then applied the similarity measures to the pattern
recognition, the clustering analysis and the medical diag-
nosis. In order to calculating the similarity measure of two
PFSs, Firozja et al. [46] presented a formula by using an
S-norm. Zhang [34] developed a new DM method based on
the similarity measures to address MCDM problems within
Pythagorean fuzzy environment based on the PFNs. Wei and
Wei [27] presented ten similarity measures between the PFSs
based on the cosine function by considering the MD, the ND
and the HD. Then, they applied these similarity measures and
weighted similarity measures between the PFSs to the pattern
recognition and the medical diagnosis.
The existing similarity measures generally have the fol-
lowing two problems: (1) The existing similarity measures
are difficult to distinguish the highly similar PFSs, which
largely limits the application range of the similarity mea-
sures; (2) In practical applications, the existing similarity
measures calculation results are prone to errors, so it is hard to
obtain good application benefits. In order to solve the above
two problems, this paper proposes several new similarity
measures based on the exponential function, and verifies that
the new similarity measures can effectively solve the above
two problems by the comparative example and the simulation
cases.
The rest of this paper is organized as follows:
Section 2 introduces the definitions of the IFS and the PFS.
Section 3 first introduces six existing similarity measures
of the PFS, then proposes four new similarity measures of
the PFS and discusses their properties, and further defines
the weighted similarity measures. Section 4 compares the
new similarity measures with the existing similarity measure
by an example. Section 5 applies the similarity measures
and the weighted similarity measures to the pattern recog-
nition, the medical diagnosis and the MCDM, respectively.
Section 6 summarizes the paper.
II. PRELIMINARIES
In this section, the related definitions of the IFS and the PFS
are introduced.
A. INTUITIONISTIC FUZZY SET
Definition 1 [47]: An IFS on the domain X is a combina-
tion of the following form
I = {hx, µI (x), νI (x)i |x ∈ X } (1)
138193
 Q. Zhang et al.: New Similarity Measures of Pythagorean FSs and Their Applications

Definition 6 [49]: Let p1 , p2 , · · · , pn be n PFNs, where

pi = (µpi , νpi ) (i = 1, 2, · · · , n). w1 , w2 , · · · , wn are the
weights corresponding to the PFNs, satisfying 0 ≤ wi ≤ 1
n
P
and wi = 1, then the Pythagorean fuzzy weighted average
i=1
(PFWA) operator is
Xn n
X
PFWA(p1 , p2 , · · · , pn ) = ( wi µpi , wi νpi ) (5)
i=1 i=1
Definition 7 [34]: Let p1 = (µp1 , νp1 ) and p2 = (µp2 , νp2 )
be two PFNs, then we can get the following ranking rules
FIGURE 1. The comparison between the IFN and the PFN.
(1) if µp1 > µp2 and νp1 < νp2 , then p1 > p2
(2) if µp1 < µp2 and νp1 > νp2 , then p1 < p2
(3) if µp1 = µp2 and νp1 = νp2 , then p1 = p2
where µI (x) : X → [0, 1] and νI (x) : X → [0, 1]. If it is For p1 = (µp1 , νp1 ) and p2 = (µp2 , νp2 ), when µp1 > µp2
satisfy for any x ∈ X , µI (x) + νI (x) ≤ 1, then µI (x) and and νp1 > νp2 (or µp1 < µp2 and νp1 < νp2 ), the above
νI (x) are respectively referred to as the MD and the ND of x ranking rules cannot rank the PFNs. Thus, we need to define
to I . another reliable ranking method.
Definition 2 [47]: For any IFS on the domain X , πI (x) is Before defining another ranking method, we need to extend
called the HD of x to I . πI (x) can be calculated as the definition of the PFS. If µP (x) and νP (x) are regarded
as vectors on the horizontal axis and the vertical axis in
πI (x) = 1 − µI (x) − νI (x) (2) the
p two-dimensional Cartesian coordinate system, rP (x) =
In addition, the intuitionistic fuzzy number (IFN) is repre- (µP (x))2 + (νP (x))2 can be regarded as the modulus of
sented by i = (µi , νi ) [48]. the sum of µP (x) and νP (x). Let the angle between rP (x)
and µP (x) be θP (x), then µP (x) = rP (x) cos(θP (x)) and
B. PYTHAGOREAN FUZZY SET νP (x) = rP (x) sin(θP (x)) can be obtained. Yager and Abbasov
Definition 3 [49]: A PFS on the domain X is a combination [2] defines a vector dP (x) that
 depicts the direction of rP (x),
of the following form and gets dP (x) = 1 − 2θp (x) π and dP (x) ∈ [0, 1].
Definition 8 [2]: Let p = (rp , dp ) be a PFN, the rank
P = {hx, µP (x), νP (x)i |x ∈ X } (3) function of p can be defined as
where µP (x) : X → [0, 1] and νP (x) : X → [0, 1]. If it is 1 1 1 1 2θp
R(p) = + rp (dp − ) = + rp ( − ) (6)
satisfy for any x ∈ X , (µP (x))2 + (νP (x))2 ≤ 1, then µP (x) 2 2 2 2 π
and νP (x) are respectively referred to as the MD and the ND For two PFNs p1 and p2 , we get can the following ranking
of x to P. rules
Definition 4 [49]: For any PFS on the domain X , πP (x) is (1) if R (p1 ) > R (p2 ), then p1 > p2
called the HD of x to P. πP (x) can be calculated as (2) if R (p1 ) < R (p2 ), then p1 < p2
(3) if R (p1 ) = R (p2 ), then p1 = p2
q
πP (x) = 1 − (µP (x))2 − (νP (x))2 (4)
This method first calculates the ranking function value,
In addition, the PFN is represented by p = (µp , νp ) [49]. and then completes the ranking of the PFNs according to the
The comparison between the IFN and the PFN is shown ranking rules.
in Figure 1.
It can be seen from Figure 1 that the IFN must be a PFN, III. SIMILARITY MEASURES
but the PFN is not necessarily an IFN. In this section, first, the existing similarity measures of the
Definition 5 [49]: Let p = (µp , νp ), p1 = (µp1 , νp1 ) and PFS are introduced. Then, the new similarity measures of the
p2 = (µp2 , νp2 ) be three PFNs, then they have the following PFS are proposed. Finally, the weighted similarity measures
operations are defined.
(1) p1 ∪ p2 = (max{µp1 , µp2 }, min{νp1 , νp2 })
(2) p1 ∩ p2 = (min{µp1 , µp2 }, max{νp1 , νp2 }) A. EXISTING SIMILARITY MEASURES
(3) pc = (νp , µq
p)  In [27], several cosine similarity measures of the PFS are
(4) p1 ⊕ p2 = µ2p1 + µ2p2 − µ2p1 µ2p2 , νp1 νp2 proposed. These similarity measures can well describe the
 q  similarity degree between the PFSs. In this paper, six typical
(5) p1 ⊗ p2 = µp1 µp2 , νp21 + νp22 − νp21 νp22 cosine similarity measures are selected from [27], which are
q 
(6) λp = 1 − (1 − µ2p )λ , (νp )λ , λ > 0 defined as follows
Definition
9 [27]: Let A =   xj , µA (x j ), νA (xj ) xj ∈ X


 q  
(7) pλ = (µp )λ , 1 − (1 − νp2 )λ , λ > 0 and B = xj , µB (xj ), νB (xj ) xj ∈ X be two PFSs on

138194 VOLUME 7, 2019

Q. Zhang et al.: New Similarity Measures of Pythagorean FSs and Their Applications

the domain X , then the six cosine similarity measures sm9 (A, B)
between A and B can be calculated as follows n
1X
=
n
sm1 (A, B) j=1
h i
µ (xj )−µ2 (xj )∨ν 2 (xj )−ν 2 (xj )∨π 2 (xj )−π 2 (xj )
 2     

1X
n
µ2 (xj )µ2B (xj ) + νA2 (xj )νB2 (xj ) × 21− A B A B A B −1
= q A q (7)
n µ4 (x ) + ν 4 (x ) µ4 (x ) + ν 4 (x ) (15)
j=1 A j A j B j B j
sm10 (A, B)
sm2 (A, B) n
n 1X
1X =
= n
n j=1
j=1 h i
1 µ (xj )−µ2 (xj )+ν 2 (xj )−ν 2 (xj )+π 2 (xj )−π 2 (xj )
 2     

µ2A (xj )µ2B (xj ) + νA2 (xj )νB2 (xj ) + πA2 (xj )πB2 (xj ) × 21− 2 A B A B A B −1
×q q (16)
µ4A (xj ) + νA4 (xj ) + πA4 (xj ) µ4B (xj ) + νB4 (xj ) + πB4 (xj )
(8)
3
where the symbol ‘‘∨’’ represents the maximum
sm (A, B) operation.
n
π µA (xj ) − µ2B (xj ) 
  2  
1X The calculation processes of the first two new similarity
= cos (9)
n 2 ∨ νA2 (xj ) − νB2 (xj ) measures are relatively simple. The latter two new similarity
j=1
measures take into account the HD factor in the calculation
sm4 (A, B) formula, so the calculation processes are slightly more com-
n
π µA (xj ) − µ2B (xj ) 
  2  
1X plicated than the first two, but the calculation results are more
= cos (10)
n 4 + νA2 (xj ) − νB2 (xj ) reliable.
j=1 The above similarity measures of the PFS are uniformly
sm5 (A, B) represented by smα ( )(α = 1, 2, · · · , 10), where the first six
µ (xj ) − µ2 (xj )
 2
  
n are the existing similarity measures, and the last four are the
1X π A B
cos   ∨ νA2 (xj ) − νB2 (xj ) 
 
= (11) new similarity measures.
n 2
j=1 ∨ πA2 (xj ) − πB2 (xj ) The main difference between the new similarity measures
and the existing similarity measures is that the function used
sm6 (A, B)
µ (xj ) − µ2 (xj )
  2   in the calculation formula is different. The existing similarity
n
1X π A 2 B measures are based on the cosine function, and the new sim-
cos   + νA (xj ) − νB2 (xj ) 

= (12) ilarity measures are based on the exponential function. The
n 4
j=1 + πA2 (xj ) − πB2 (xj ) different functions will have a direct impact on the calculation
results of the similarity measures.
where the symbol ‘‘∨’’ represents the maximum operation. For the two PFSs A and B on the domain X , their similarity
measures should satisfy the following properties.
B. NEW SIMILARITY MEASURES Property 1: 0 ≤ smα (A, B) ≤ 1.
In this paper, several new similarity measures based on the Proof (Take sm7 (A, B) as an Example): Prove that
exponential function are proposed, which are defined as 0 ≤ sm7 (A,  B) ≤ 
1 needs to prove 0 ≤
1− µ2A (xj )−µ2B (xj )∨νA2 (xj )−νB2 (xj )

follows 2 − 1 ≤ 1. Let t =
Definition
10: Let A =   xj , µA (x j ), νA (xj ) xj ∈ X

µ2 (xj ) − µ2 (xj ) ∨ ν 2 (xj ) − ν 2 (xj ) and y = 21−t − 1,

  

A B A B
and B = xj , µB (xj ), νB (xj ) q
xj ∈ X be two PFSs on the function curve of y is shown in Figure 2. It can be seen
the domain X , and πA (xj ) = 1 − (µA (xj ))2 − (νA (xj ))2 , from Figure 2 that 0 ≤ y ≤ 1, thus 0 ≤ sm7 (A, B) ≤ 1.
q Property 2: smα (A, B) = 1 if and only if A = B.
πB (xj ) = 1 − (µB (xj ))2 − (νB (xj ))2 , then the four similarity Proof (Take sm8 (A, B) as an Example): For two PFSs A
measures between A and B can be calculated as follows and B on the domain X , if A = B, then µ2A (xj ) = µ2B (xj ) and
νA2 (xj ) = νB2 (xj ). Obviously, µ2A (xj ) − µ2B (xj ) = 0, νA2 (xj ) −
sm7 (A, B) νB2 (xj ) = 0 can be obtained, 8
 thus sm (A, B)  = 1. 
n
1 X h 1− µ (xj )−µ2 (xj )∨ν 2 (xj )−ν 2 (xj )
 2   
i If sm 8 (A, B) = 1, then µ2A (xj ) − µ2B (xj ) = 0 and νA2 (xj ) −
=
n
2 A B A B −1 (13) νB2 (xj ) = 0are satisfied. Obviously, µ2A (xj ) = µ2B (xj ) and
j=1 νA2 (xj ) = νB2 (xj ) can be obtained, thus A = B.
sm8 (A, B) Property 3: smα (A, B) = smα (B, A).
n
1 X h 1− 1 µ (xj )−µ2 (xj )+ν 2 (xj )−ν 2 (xj )
 2   
i Proof (Take sm9 (A, B) as an Example): Prove that
9 (B, A) needs to prove µ2 (x ) − µ2 (x ) =
9

= 2 2 A B A B −1 (14) sm (A, B) = sm A j B j
n µ (xj ) − µ2 (xj ), ν 2 (xj )− ν 2 (xj ) = ν 2 (xj ) − ν 2 (xj ) and
 2    
j=1
B A A B B A

VOLUME 7, 2019 138195

 Q. Zhang et al.: New Similarity Measures of Pythagorean FSs and Their Applications

sm2w (A, B)
n
X
= wj
j=1

µ2A (xj )µ2B (xj ) + νA2 (xj )νB2 (xj ) + πA2 (xj )πB2 (xj )
×q q
µ4A (xj ) + νA4 (xj ) + πA4 (xj ) µ4B (xj ) + νB4 (xj ) + πB4 (xj )
(18)
sm3w (A, B)
n µ (xj ) − µ2 (xj )
"  2  !#
X π A 2 B
= wj cos (19)
∨ ν (xj ) − ν 2 (xj )

FIGURE 2. Function curve of y. 2 A B
j=1

π (xj )− π 2 (xj ) = π 2 (xj ) − π 2 (xj ). Obviously, this is

 2    sm4w (A, B)
A B B A n µ (xj ) − µ2 (xj )
"  2  !#
true, so it is proven. X π A B
= wj cos (20)
+ ν 2 (xj ) − ν 2 (xj )
 
Property 4: If C is a PFS on the domain X and A ⊆ B ⊆ C 4 A B
then smα (A, C) ≤ smα (A, B), smα (A, C) ≤ smα (B, C).
j=1

Proof (Take sm10 (A, B) as an Example): If A ⊆ B ⊆ C, sm5w (A, B)

then µA (xj ) ≤ µB (xj ) ≤ µC (xj ), νA (xj ) ≥ νB (xj ) ≥ νC (xj ), µ (xj ) − µ2 (xj )
 2
  
n A B
πA (xj ) ≤ πB (xj ) ≤ πC (xj ), and then µ2A (xj ) ≤ µ2B (xj ) ≤ π 
wj cos   ∨ νA2 (xj ) − νB2 (xj ) 
X   
µ2C (xj ), νA2 (xj ) ≥ νB2 (xj ) ≥ νC2 (xj ), πA2 (xj ) ≤ πB2 (xj ) ≤ πC2 (xj ), = (21)
2
∨ π 2 (xj ) − π 2 (xj )
 
j=1
thus A B
    sm6w (A, B)
µA (xj ) − µ2B (xj ) ≤ µ2A (xj ) − µ2C (xj )
 2   
µ (xj ) − µ2 (xj )
 2
  
n A B
π  
   
µB (xj ) − µ2C (xj ) ≤ µ2A (xj ) − µ2C (xj )
 2
wj cos   + νA2 (xj ) − νB2 (xj ) 
   X  
= (22)
4
+ π 2 (xj ) − π 2 (xj )
 
    j=1
A B
νA (xj ) − νB2 (xj ) ≤ νA2 (xj ) − νC2 (xj )
 2   
    sm7w (A, B)
νB (xj ) − νC2 (xj ) ≤ νA2 (xj ) − νC2 (xj )
 2   
n h i
µ (xj )−µ2 (xj )∨ν 2 (xj )−ν 2 (xj )
X  2   

    = wj 21− A B A B −1 (23)
πA (xj ) − πB2 (xj ) ≤ πA2 (xj ) − πC2 (xj )
 2   
j=1

sm8w (A, B)
   
πB (xj ) − πC2 (xj ) ≤ πA2 (xj ) − πC2 (xj )
 2   
n h i
1 µ (xj )−µ2 (xj )+ν 2 (xj )−ν 2 (xj )
X  2   

= wj 21− 2 A B A B −1 (24)
So sm10 (A, C) ≤ sm10 (A, B)
and sm10 (A, C)
≤
j=1
sm10 (B, C).
Considering that in the MCDM, decision criteria will be sm9w (A, B)
given different weights depending on the importance degree. n
X
Therefore, the criteria weights need to be considered when = wj
calculating the similarity measures of the PFS. Next, the cri- j=1
teria weights are introduced into the similarity measures to h
µ (xj )−µ2 (xj )∨ν 2 (xj )−ν 2 (xj )∨π 2 (xj )−π 2 (xj )
 2     
i
obtain the weighted similarity measures, which are defined × 21− A B A B A B −1
as follows (25)
11: Let A = xj , µ A (xj ), νA (xj ) xj ∈ X and


Definition
sm10
w (A, B)
B = xj , µB (xj ), νB (xj ) xj ∈ X be two PFS on the domain


n
X , w = (w1 , w2 , · · · , wn )T be the weights corresponding to X
= wj
the decision criteria. Then the weighted similarity measures
j=1
between A and B can be calculated as follows h i
1 µ (xj )−µ2 (xj )+ν 2 (xj )−ν 2 (xj )+π 2 (xj )−π 2 (xj )
 2     

× 21− 2 A B A B A B −1
sm1w (A, B)
n
(26)
X µ2 (xj )µ2B (xj ) + νA2 (xj )νB2 (xj )
= wj q A q (17) The above weighted similarity measures of the PFS are
j=1 µ4A (xj ) + νA4 (xj ) µ4B (xj ) + νB4 (xj ) β
uniformly represented by smw ( )(β = 1, 2, · · · , 10), where

138196 VOLUME 7, 2019

Q. Zhang et al.: New Similarity Measures of Pythagorean FSs and Their Applications
TABLE 1. Calculation results of smα (A, Ai ).
the first six are the existing weighted similarity measures, and
the last four are the new weighted similarity measures.
IV. COMPARATIVE EXAMPLE
In this section, the new similarity measures are compared
with the existing similarity measures by an example, and the
advantages of the new similarity measures are verified by
comparison with the existing similarity measures, and the
weighted similarity measures are also compared.
It is generally believed that the excellent similarity mea-
sures must have good distinguishability. Next, for the highly
similar but inconsistent PFSs, the similarity measures and the
weighted similarity measures are used to distinguish them,
respectively.
Let the feature space be X = {x1 , x2 , x3 }, there are three
PFSs on X , where
A1 = {(x1 , 0.5, 0.31), (x2 , 0.6, 0.2), (x3 , 0.8, 0.1)}
A2 = {(x1 , 0.5, 0.3), (x2 , 0.6, 0.22), (x3 , 0.8, 0.1)}
A3 = {(x1 , 0.5, 0.3), (x2 , 0.6, 0.2), (x3 , 0.8, 0.13)}
There is a PFS A = {(x1 , 0.5, 0.3), (x2 , 0.6, 0.2), (x3 , 0.8,
0.1)} that is highly similar but not exactly the same as A1 , A2
and A3 . How to distinguish A from A1 , A2 and A3 is the next
step to be studied. Thus, the existing similarity measures and
the new similarity measures between A and A1 , A2 , A3 are
calculated. The calculation results of the similarity measures
|xi ∈ X } (i = 1, 2,
TABLE 2. Calculation results of smαw (A, Ai ).
A1 , A2 , A3 . In this case, the existing weighted similarity
measures are unreasonable, but the new weighted similarity
measures can effectively distinguish them.
In summary, for the highly similar but inconsistent PFSs,
the existing similarity measures and the existing weighted
similarity measures cannot distinguish them, while the new
similarity measures and the new weighted similarity mea-
sures can effectively distinguish them. So, the new similarity
measures and the new weighted similarity measures have
better distinguishability and the wider application range.
V. SIMULATION CASE
In this section, the similarity measures and the weighted
similarity measures are applied in three cases. These three
cases are about the pattern recognition, the medical diagnosis
and the MCDM, respectively. We can test the performance
of the similarity measures in dealing with different practical
application problems, and then can verify the superiority
of the new similarity measures compared with the existing
similarity measures, and verify the superiority of the new
weighted similarity measures.
A. A CASE OF PATTERN RECOGNITION
Suppose there are m patterns represented by the PFSs,
Ai = {hxi , µA (xi ), νA (xi )i
 · · · , m),
and
are shown in Table 1.
From the Table 1, it can be seen that the existing sim-
ilarity measures between A and A1 , A2 , A3 all are 1.0000
(except sm1 ( )), which cannot distinguish A from A1 , A2 , A3 .
In this case, the existing similarity measures are unreason-
able, but the new similarity measures can effectively distin-
guish them.
When the weight of the feature space X is w =
(0.2, 0.4, 0.4), the existing weighted similarity measures and
the new weighted similarity measures between A and A1 , A2 ,
A3 are calculated. The calculation results of the similarity
measures are shown in Table 2.
From the Table 2, it can be seen that the existing
weighted similarity measures also cannot distinguish A from
VOLUME 7, 2019
an unknown pattern B = xj , µB (xj ), νB (xj ) xj ∈ X rep-
resented by the PFSs needs to be recognized. The pattern
recognition is to classify B in one of classes Ai .
Let the feature space be X = {x1 , x2 , x3 }, there are three
known patterns which are represented by the PFSs, where
A1 = {(x1 , 1.0, 0.0), (x2 , 0.8, 0.0), (x3 , 0.7, 0.1)}
A2 = {(x1 , 0.8, 0.1), (x2 , 1.0, 0.0), (x3 , 0.9, 0.1)}
A3 = {(x1 , 0.6, 0.2), (x2 , 0.8, 0.0), (x3 , 1.0, 0.0)}
There is an unknown pattern B needs to be recognized,
where
B = {(x1 , 0.5, 0.3), (x2 , 0.6, 0.2), (x3 , 0.8, 0.1)}
138197
 Q. Zhang et al.: New Similarity Measures of Pythagorean FSs and Their Applications

TABLE 3. Calculation results of smα (B, Ai ). calculated by (5).

PFWA(A1 ) = (0.8333, 0.0333)
PFWA(A2 ) = (0.9000, 0.0667)
PFWA(A3 ) = (0.8000, 0.0667)
PFWA(B) = (0.6333, 0.2000)
Then, the ranking function value of the comprehensive
assessment value of each pattern is calculated by (6).
R(A1 ) = 0.8958
R(A2 ) = 0.9087
R(A3 ) = 0.8589
R(B) = 0.7027
According to the size of the ranking function value, R(B)
TABLE 4. Ranking results of smα (B, Ai ). and R(A3 ) are the closest, so B and A3 belong to the same
pattern. The known patterns can be ranked according to the
difference between the rank function value of the known
patterns and the unknown pattern.
V (A1 B) = R(A1 ) − R(B) = 0.1931
V (A2 B) = R(A2 ) − R(B) = 0.2060
V (A3 B) = R(A3 ) − R(B) = 0.1562
Since V (A3 B) < V (A1 B) < V (A2 B), the ranking result
is A3 > A1 > A2 , which is completely consistent with
the ranking results obtained by the new similarity measures.
Compared with the existing similarity measures, the new
similarity measures are more accurate, thus the application
process is more reliable.

B. A CASE OF MEDICAL DIAGNOSIS

In order to determine which pattern B is, the existing
similarity measures and the new similarity measures between
B and A1 , A2 , A3 are calculated, respectively. The calculation
results of the similarity measures are shown in Table 3.
Next, the similarity measures between the three known
patterns and the unknown pattern are ranked, and the ranking
results are shown in Table 4.
It can be seen from Table 4 that the new similarity measures
and the existing similarity measures between B and A3 are the
largest, so B and A3 belong to the same pattern. Moreover,
the ranking results obtained by the new similarity measures
all are A3 > A1 > A2 , and the ranking results obtained by the
existing similarity measures all are A3 > A2 > A1 .
The ranking results obtained by the new similarity mea-
sures and the existing similarity measures are different.
In order to verify the accuracy of the new similarity measures,
the PFWA operator in Definition 6 and the ranking method
in Definition 8 (hereinafter referred to as the verification
method) are used to rank the known patterns.
Considering that the weight of the three features in the
feature space is not given, that is, the importance
 ofthe three
 
features is the same, thus the weight is w = (1 3, 1 3, 1 3).
First, the comprehensive assessment value of each pattern is
There are many diseases, different diseases show different
symptoms. The medical diagnosis is based on the patient’s
symptoms to determine what kind of disease the patient has
suffered. The patient’s multiple symptoms can constitute a
symptom set, and different diseases can constitute a diagnos-
tic set.
Let the symptom set be S = {s1 (Temperature),
s2 (Headache)s3 (Stomachpain), s4 (Cough), s5 (Chestpain)},
and the diagnostic set be D = {D1 (Viral fever), D2 (Malaria),
D3 (Typhoid), D4 (Stomachproblem), D5 (Chestproblem)}. All
symptoms of a patient can be represented by the PFSs as
follows
(s1 , 0.8, 0.1), (s2 , 0.6, 0.1), (s3 , 0.2, 0.8),
 
P(patiment) =
(s4 , 0.6, 0.1), (s5, 0.1, 0.6)
The symptoms of each disease Di (i = 1, 2, 3, 4, 5) can be
expressed by the PFSs as follows
D1 (Viral fever)
(s1 , 0.4, 0.0), (s2 , 0.3, 0.5), (s3 , 0.1, 0.7),
 
=
(s4 , 0.4, 0.3), (s5, 0.1, 0.7)
D2 (Malaria)
(s1 , 0.7, 0.0), (s2 , 0.2, 0.6), (s3 , 0.0, 0.9),

=
(s4 , 0.7, 0.0), (s5, 0.1, 0.8)

138198 VOLUME 7, 2019

Q. Zhang et al.: New Similarity Measures of Pythagorean FSs and Their Applications

β TABLE 7. Pythagorean fuzzy decision matrix constructed by e1 .

TABLE 5. Calculation results of smw (P, Di ).

TABLE 8. Pythagorean fuzzy decision matrix constructed by e2 .

β
TABLE 6. Ranking results of smw (P, Di ).

TABLE 9. Pythagorean fuzzy decision matrix constructed by e3 .

Considering that the ranking results of the existing

D3 (Typhoid) weighted similarity measures are not completely consistent

(s1 , 0.3, 0.3), (s2 , 0.6, 0.1), (s3 , 0.2, 0.7),
 with those of the new weighted similarity measures, the dis-
= eases Di are ranked by the verification method.
(s4 , 0.2, 0.6), (s5, 0.1, 0.9)
First, the comprehensive assessment value of P and Di is
D4 (Stomach problem)
calculated by (5).
(s1 , 0.1, 0.7), (s2 , 0.2, 0.4), (s3 , 0.8, 0.0),
 
=
(s4 , 0.2, 0.7), (s5, 0.2, 0.7) PFWA(P) = (0.4550, 0.3300)
D5 (Chest problem) PFWA(D1 ) = (0.2550, 0.4500)
(s1 , 0.1, 0.8), (s2 , 0.0, 0.8), (s3 , 0.2, 0.8),
 
= PFWA(D2 ) = (0.3200, 0.4850)
(s4 , 0.2, 0.8), (s5, 0.8, 0.1)
PFWA(D3 ) = (0.2950, 0.5050)
If the weight of the symptom set is w = (0.20, 0.25, PFWA(D4 ) = (0.2700, 0.5200)
0.15, 0.15, 0.25), then the weighted similarity measures
PFWA(D5 ) = (0.2800, 0.6250)
between P and Di (i = 1, 2, · · · , 5) are calculated. The
calculation results are shown in Table 5. Then, the ranking function value of the comprehensive
Next, the weighted similarity measures between P and assessment value of each pattern is calculated by (6).
Di (i = 1, 2, · · · , 5) are ranked, and the ranking results are
shown in Table 6. R(P) = 0.5565
It can be seen from Table 6 that the ranking results obtained R(D1 ) = 0.4111
by the weighted similarity measures are D2 > D1 > D3 > R(D2 ) = 0.4252
D4 > D5 (except sm1w ( ) and sm4w ( )). The new weighted simi-
R(D3 ) = 0.4044
larity measures and the existing weighted similarity measures
between P and D2 is the largest, so the patient’s disease is R(D4 ) = 0.3857
diagnosed as malaria. R(D5 ) = 0.3412

VOLUME 7, 2019 138199

 Q. Zhang et al.: New Similarity Measures of Pythagorean FSs and Their Applications

TABLE 10. Comprehensive pythagorean fuzzy decision matrix.

According to the size of the ranking function value, R(P) β

and R(D2 ) are the closest, so P and D2 belong to the same dis-
ease, namely malaria. The diseases can be ranked according
to the difference between the rank function value of P and Di .
V (PD1 ) = R(P) − R(D1 ) = 0.1454
V (PD2 ) = R(P) − R(D2 ) = 0.1313
V (PD3 ) = R(P) − R(D3 ) = 0.1521
V (PD4 ) = R(P) − R(D4 ) = 0.1708
V (PD5 ) = R(P) − R(D5 ) = 0.2153
Since V (PD2 ) < V (PD1 ) < V (PD3 ) < V (PD4 ) <
V (PD5 ), the ranking result is D2 > D1 > D3 > D4 >
D5 , which is completely consistent with the ranking results
obtained by the new weighted similarity measures. The rank-
ing results obtained by the two existing weighted similarity
measures (sm1w ( ) and sm4w ( )) are not accurate, which show
that the new weighted similarity measures are more accurate
and reliable.
C. A CASE OF MULTIPLE CRITERIA DECISION MAKING
The MCDM problem can be described as follows: let A =
{A1 , A2 , · · · , Am }(i = 1, 2, · · · , m) be m decision schemes,
C = {C1 , C2 , · · · , Cn }(j = 1, 2, · · · , n) be n decision
criteria, E = {e1 , e2 , · · · , el }(k = 1, 2, · · · , l) be l decision
makers. Each decision maker judges each scheme according
to n criteria, so as to rank all schemes optimally.
An organization plans to implement enterprise resource
planning (ERP) system. There are five potential ERP systems
as candidates, and the scheme set is A = {A1 , A2 , A3 , A4 , A5 }.
The organization employs three experts as decision makers,
and the expert set is E = {e1 , e2 , e3 }. The organization
selects four criteria to assess candidate schemes: functional
and technical C1 , strategic adaptability C2 , supplier capability
C3 , supplier reputation C4 , then the criterion set is C =
{C1 , C2 , C3 , C4 }. The three experts ek (k = 1, 2, 3) assess
the five potential ERP systems Ai (i = 1, 2, 3, 4, 5) under the
four criteria Cj (j = 1, 2, 3, 4) .The weight of the expert set
is v = (0.25, 0.45, 0.30), the weight of the criterion set is
w = (0.15, 0.20, 0.35, 0.30).
Next, each expert’s Pythagorean fuzzy decision matrix
(k)
P(k) = (pij )5×4 is constructed, as shown in Table 7-Table 9,
respectively.
By using (5) to aggregate the Pythagorean fuzzy decision
(k)
matrix P(k) = (pij )5×4 of each decision maker to obtain
TABLE 11. Calculation results of smw (pA , p+ ).
i
the comprehensive Pythagorean fuzzy decision matrix P =
(pij )5×4 , as shown in Table 10.
Considering that four criteria all are benefit index, then the
positive ideal solution p+ can be defined as
p+ = p+ 1 , p2 , p3 , p4 = {(1, 0), (1, 0), (1, 0), (1, 0)}
 + + +
β
Next, the weighted similarity measures smw (pAi , p+ )
between the each scheme’s criterion value vector pAi =
{pi1 , pi2 , pi3 , pi4 } and the positive ideal solution p+ are cal-
culated. The calculation results are shown in Table 11.
Next, the weighted similarity measures between pAi and p+
are ranked, and the ranking results are shown in Table 12.
It can be seen from Table 12 that the ranking results
obtained by the weighted similarity measures are A2 > A3 >
A4 > A1 > A5 (except sm1w ( )). The new weighted similar-
ity measures and the existing weighted similarity measures
between pA2 and p+ is the largest, so the best ERP system
is A2 .
Considering that the ranking results obtained by the exist-
ing weighted similarity measures are not completely con-
sistent with those obtained by the new weighted similarity
measures, the five potential ERP systems Ai are ranked by
the verification method.
First, the comprehensive assessment value of Ai is calcu-
lated by (5).
PFWA(A1 ) = (0.5130, 0.3155)
PFWA(A2 ) = (0.6850, 0.3375)

138200 VOLUME 7, 2019

Q. Zhang et al.: New Similarity Measures of Pythagorean FSs and Their Applications
β (3) The simulation case of the pattern recognition verifies
TABLE 12. Ranking results of smw (pA , p+ ).
i
PFWA(A3 ) = (0.6148, 0.3850)
PFWA(A4 ) = (0.5710, 0.3902)
PFWA(A5 ) = (0.4688, 0.2435)
Then, the ranking function value of the comprehensive
assessment value of each pattern is calculated by (6).
R(A1 ) = 0.5897
R(A2 ) = 0.6593
R(A3 ) = 0.6043
R(A4 ) = 0.5819
R(A5 ) = 0.6030
According to the ranking rules in Definition 8, the ranking
result of all schemes is A2 > A3 > A4 > A1 > A5 , which
is completely consistent with the ranking results obtained by
the new weighted similarity measures. Thus, the accuracy
and reliability of the new weighted similarity measures are
verified.
VI. CONCLUSION
In this paper, four new similarity measures of the PFS are
proposed, and the new weighted similarity measures are
obtained. By a comparative example and three simulation
cases, it is verified that the new similarity measures can
effectively overcome the limitations of the existing similarity
measures. The main conclusions are as follows:
(1) The comparative example verifies that the new sim-
ilarity measures and the new weighted similarity measures
can effectively distinguish the highly similar PFSs, so the
new similarity measures and the new weighted similarity
measures have better distinguishability.
(2) The simulation cases verify that the new similarity
measures and the new weighted similarity measures can
effectively solve the application problems of the pattern
recognition, the medical diagnosis and the MCDM, so the
new similarity measures and the new weighted similarity
measures can achieve good application benefits.
VOLUME 7, 2019
that the new similarity measures are more accurate and reli-
able than the existing similarity measures.
(4) The simulation cases of the medical diagnosis and the
MCDM verify that the new weighted similarity measures
are more accurate and reliable than the existing weighted
similarity measures.
The PFS has been extended, such as the Pythagorean fuzzy
linguistic set. The next step of our work is to study the
similarity measures of the extended PFS.
REFERENCES
[1] K. T. Atanassov, ‘‘Intuitionistic fuzzy sets,’’ Fuzzy Sets Syst., vol. 20,
pp. 87–96, Aug. 1986.
[2] R. R. Yager and A. M. Abbasov, ‘‘Pythagorean membership grades, com-
plex numbers, and decision making,’’ Int. J. Intell. Syst., vol. 28, no. 5,
pp. 436–452, 2013.
[3] R. R. Yager, ‘‘Pythagorean fuzzy subsets,’’ in Proc. Joint IFSA World
Congr. NAFIPS Annu. Meeting, Edmonton, AB, Canada, 2013, pp. 57–61.
[4] P. A. Ejegwa, ‘‘Pythagorean fuzzy set and its application in career place-
ments based on academic performance using max–min–max composi-
tion,’’ Complex Intell. Syst., vol. 5, no. 2, pp. 165–175, 2019.
[5] X. Peng and Y. Yang, ‘‘Some results for Pythagorean fuzzy sets,’’ Int.
J. Intell. Syst., vol. 30, no. 11, pp. 1133–1160, 2015.
[6] X. Peng and H. Yuan, ‘‘Fundamental properties of pythagorean fuzzy
aggregation operators,’’ Fundam. Inform., vol. 147, no. 4, pp. 415–446,
2016.
[7] H. Garg, ‘‘A novel correlation coefficients between Pythagorean fuzzy sets
and its applications to decision-making processes,’’ Int. J. Intell. Syst.,
vol. 12, no. 3, pp. 1–19, 2016.
[8] R. R. Yager, ‘‘Pythagorean membership grades in multicriteria decision
making,’’ IEEE Trans. Fuzzy Syst., vol. 22, no. 4, pp. 958–965, Aug. 2014.
[9] X. D. Peng and G. Selvachandran, ‘‘Pythagorean fuzzy set: State of the
art and future directions,’’ Artif. Intell. Rev., vol. 52, no. 3, pp. 1873–1927,
2017.
[10] X. Peng, ‘‘Algorithm for pythagorean fuzzy multi-criteria decision making
based on WDBA with new score function,’’ Fundamenta Informaticae,
vol. 165, no. 3, pp. 99–137, 2019.
[11] X. Peng and W. Li, ‘‘Algorithms for interval-valued pythagorean fuzzy
sets in emergency decision making based on multiparametric similarity
measures and WDBA,’’ IEEE Access, vol. 7, pp. 7419–7441, 2019.
[12] M. Lu, G. Wei, F. E. Alsaadi, T. Hayat, and A. Alsaedi, ‘‘Hesitant
pythagorean fuzzy Hamacher aggregation operators and their application
to multiple attribute decision making,’’ J. Intell. Fuzzy Syst., vol. 33, no. 2,
pp. 1105–1117, 2017.
[13] H. Garg, ‘‘Linguistic Pythagorean fuzzy sets and its applications in mul-
tiattribute decision-making process,’’ Int. J. Intell. Syst., vol. 33, no. 6,
pp. 1234–1263, 2018.
[14] C. Zhang and D. Y. Li, ‘‘Pythagorean fuzzy rough sets and its applications
in multi-attribute decision making,’’ J. Chin. Comput. Syst., vol. 37, no. 3,
pp. 1531–1535, 2016.
[15] C.-M. Hwang, M.-S. Yang, W.-L. Hung, and M.-G. Lee, ‘‘A similarity
measure of intuitionistic fuzzy sets based on the Sugeno integral with
its application to pattern recognition,’’ Inf. Sci., vol. 189, pp. 93–109,
Apr. 2012.
[16] X. D. Peng and H. Garg, ‘‘Multiparametric similarity measures on
Pythagorean fuzzy sets with applications to pattern recognition,’’ Appl.
Intell., to be published.
[17] Y. Song, X. Wang, L. Lei, and A. Xue, ‘‘A novel similarity measure on
intuitionistic fuzzy sets with its applications,’’ Appl. Intell., vol. 42, no. 2,
pp. 252–261, Mar. 2015.
[18] Y. B. Gong, ‘‘A new similarity measures of intuitionistic fuzzy sets and
application to pattern recognitions,’’ Adv. Mater. Res., vols. 219–220,
pp. 160–164, Mar. 2011.
[19] Q. S. Zhang, H. X. Yao, and Z. H. Zhang, ‘‘Some similarity measures of
interval-valued intuitionistic fuzzy sets and application to pattern recogni-
tion,’’ Appl. Mech. Mater., vols. 44–47, pp. 3888–3892, Dec. 2010.
[20] D. Li and C. Cheng, ‘‘New similarity measures of intuitionistic fuzzy sets
and application to pattern recognitions,’’ Pattern Recognit. Lett., vol. 23,
nos. 1–3, pp. 221–225, 2002.
138201
 Q. Zhang et al.: New Similarity Measures of Pythagorean FSs and Their Applications
[21] X. Peng, H. Yuan, and Y. Yang, ‘‘Pythagorean fuzzy information measures
and their applications,’’ Int. J. Intell. Syst., vol. 32, no. 2, pp. 991–1029,
2017.
[22] P. Muthukumar and G. S. S. Krishnan, ‘‘A similarity measure of intuition-
istic fuzzy soft sets and its application in medical diagnosis,’’ Appl. Soft
Comput., vol. 41, pp. 148–156, Apr. 2016.
[23] L. H. Son and P. H. Phong, ‘‘On the performance evaluation of intuitionistic
vector similarity measures for medical diagnosis,’’ J. Intell. Fuzzy Syst.,
vol. 31, no. 3, pp. 1597–1608, 2016.
[24] C.-P. Wei, P. Wang, and Y.-Z. Zhang, ‘‘Entropy, similarity measure of
interval-valued intuitionistic fuzzy sets and their applications,’’ Inf. Sci.,
vol. 181, no. 19, pp. 4273–4286, 2011.
[25] K.-C. Hung and P.-K. Wang, ‘‘An integrated intuitionistic fuzzy similarity
measures for medical problems,’’ Int. J. Comput. Intell. Syst., vol. 7, no. 2,
pp. 327–343, 2014.
[26] T. Maoying, ‘‘A new fuzzy similarity measure based on cotangent function
for medical diagnosis,’’ Adv. Model. Optim., vol. 15, no. 3, pp. 151–156,
2013.
[27] G. Wei and Y. Wei, ‘‘Similarity measures of Pythagorean fuzzy sets based
on the cosine function and their applications,’’ Int. J. Intell. Syst., vol. 33,
no. 3, pp. 634–652, 2018.
[28] J. Ye, ‘‘Multicriteria decision-making method using the Dice similar-
ity measure between expected intervals of trapezoidal fuzzy numbers,’’
J. Decision Syst., vol. 21, no. 4, pp. 307–317, 2012.
[29] J. Ye, ‘‘Multicriteria group decision-making method using vector similarity
measures for Trapezoidal intuitionistic fuzzy numbers,’’ Group Decision
Negotiation, vol. 21, no. 3, pp. 519–530, 2012.
[30] S.-M. Chen, S.-H. Cheng, and T.-C. Lan, ‘‘Multicriteria decision making
based on the TOPSIS method and similarity measures between intuition-
istic fuzzy values,’’ Inf. Sci., vols. 367–368, pp. 279–295, Nov. 2016.
[31] J. Ye, ‘‘Interval-valued intuitionistic fuzzy cosine similarity measures for
multiple attribute decision-making,’’ Int. J. General Syst., vol. 42, no. 8,
pp. 883–891, 2013.
[32] D. Liang and Z. Xu, ‘‘The new extension of TOPSIS method for multiple
criteria decision making with hesitant Pythagorean fuzzy sets,’’ Appl. Soft
Comput., vol. 60, pp. 167–179, Nov. 2017.
[33] Z. Xu, ‘‘Some similarity measures of intuitionistic fuzzy sets and their
applications to multiple attribute decision making,’’ Fuzzy Optim. Decis.
Making, vol. 6, no. 2, pp. 109–121, Jun. 2007.
[34] X. Zhang, ‘‘A novel approach based on similarity measure for pythagorean
fuzzy multiple criteria group decision making,’’ Int. J. Intell. Syst., vol. 31,
no. 6, pp. 593–611, 2016.
[35] W.-L. Hung and M.-S. Yang, ‘‘On similarity measures between intuition-
istic fuzzy sets,’’ Int. J. Intell. Syst., vol. 23, no. 3, pp. 364–383, 2010.
[36] H.-W. Liu, ‘‘New similarity measures between intuitionistic fuzzy sets and
between elements,’’ Math. Comput. Model., vol. 42, pp. 61–70, Jul. 2005.
[37] E. Szmidt and J. Kacprzyk, ‘‘A new concept of a similarity measure for
intuitionistic fuzzy sets and its use in group decision making,’’ in Modeling
Decisions for Artificial Intelligence (Lecture Notes in Computer Science),
vol. 3558, 2005, pp. 272–282.
[38] W.-L. Hung and M.-S. Yang, ‘‘Similarity measures of intuitionistic fuzzy
sets based on Hausdorff distance,’’ Pattern Recognit. Lett., vol. 25, no. 14,
pp. 1603–1611, Oct. 2004.
[39] W.-L. Hung and M.-S. Yang, ‘‘Similarity measures of intuitionistic fuzzy
sets based on Lp metric,’’ Int. J. Approx. Reasoning, vol. 46, no. 1,
pp. 120–136, 2007.
[40] M. Xia and Z. Xu, ‘‘Some new similarity measures for intuitionistic fuzzy
values and their application in group decision making,’’ J. Syst. Sci. Syst.
Eng., vol. 19, no. 4, pp. 430–452, 2010.
[41] J. Ye, ‘‘Similarity measures of intuitionistic fuzzy sets based on cosine
function for the decision making of mechanical design schemes,’’ J. Intell.
Fuzzy Syst., vol. 30, no. 1, pp. 151–158, 2016.
[42] C.-M. Hwang, M.-S. Yang, and W.-L. Hung, ‘‘New similarity measures of
intuitionistic fuzzy sets based on the Jaccard index with its application to
clustering,’’ Int. J. Intell. Syst., vol. 33, no. 8, pp. 1672–1688, 2018.
[43] S.-M. Chen, S.-H. Cheng, and T.-C. Lan, ‘‘A novel similarity measure
between intuitionistic fuzzy sets based on the centroid points of trans-
formed fuzzy numbers with applications to pattern recognition,’’ Inf. Sci.,
vols. 343–344, pp. 15–40, May 2016.
[44] P. A. Ejegwa, ‘‘Distance and similarity measures for Pythagorean fuzzy
sets,’’ Granular Comput., to be published.
[45] W. Zeng, D. Li, and Q. Yin, ‘‘Distance and similarity measures of
Pythagorean fuzzy sets and their applications to multiple criteria group
decision making,’’ Int. J. Intell., vol. 33, no. 11, pp. 2236–2254, 2018.
138202
[46] M. A. Firozja, B. Agheli, and E. B. Jamkhaneh, ‘‘A new similarity measure
for Pythagorean fuzzy sets,’’ Complex Intell. Syst., to be published.
[47] K. T. Atanassov, ‘‘More on intuitionistic fuzzy sets,’’ Fuzzy Sets Syst.,
vol. 33, no. 1, pp. 37–45, Oct. 1989.
[48] Z. Xu, ‘‘Intuitionistic fuzzy aggregation operators,’’ IEEE Trans. Fuzzy
Syst., vol. 15, no. 6, pp. 1179–1187, Dec. 2007.
[49] X. Zhang and Z. Xu, ‘‘Extension of TOPsIs to multiple criteria decision
making with Pythagorean fuzzy sets,’’ Int. J. Intell. Syst., vol. 29, no. 12,
pp. 1061–1078, 2014.
QIANG ZHANG was born in 1991. He received
the M.S. degree from Air Force Engineering Uni-
versity, Xi’an, China, in 2016, where he is cur-
rently pursuing the Ph.D. degree. His research
interests include fuzzy set theory, intelligent
decision-making, and multi-attribute decision-
making theory.
JUNHUA HU was born in 1980. He received the
M.S. and Ph.D. degrees from Air Force Engineer-
ing University, Xi’an, China, in 2004 and 2008,
respectively, where he is currently pursuing the
Ph.D. degree. He is currently an Associate Pro-
fessor with Air Force Engineering University. His
research interests include the intelligent decision-
making of air combat and weapon management
systems.
JINFU FENG was born in 1964. He received the
Ph.D. degree from the Nanjing University of Sci-
ence and Technology, in 1996. He is currently
pursuing the Ph.D. degree with Air Force Engi-
neering University, Xi’an, China, where he is cur-
rently a Professor. His research interests include
the intelligent decision-making of air combat, air-
borne stores management systems, and design of
cross-medium weapons.
AN LIU was born in 1982. He received the
M.S. and Ph.D. degrees from Air Force Engineer-
ing University, Xi’an, China, in 2007 and 2010,
respectively, where he is currently pursuing the
Ph.D. degree. He is currently a Lecturer with Air
Force Engineering University. His research inter-
ests include the intelligent decision-making of air
combat and aviation bus management systems.
YONGLI LI was born in 1980. He received the
M.S. and Ph.D. degrees from Air Force Engineer-
ing University, in 2013 and 2017, respectively. He
is currently pursuing the Ph.D. degree with the
Engineering University of CAPF, Xi’an, China,
where he is currently a Lecturer. His research inter-
ests include the intelligent decision-making of air
combat and control of cross-medium weapons.
VOLUME 7, 2019