You are on page 1of 23

JOURNAL OF MECHANICS OF CONTINUA AND

MATHEMATICAL SCIENCES
www.journalimcms.org
0
J. Mech.
ISSN (Online) : 2454 -7190 Cont.
Vol.-16, & Math.
No.-6, Sci., Vol.-16,
June (2021) pp 64-86No.-6,
ISSN June
(Print)(2021) pp 64-86
0973-8975

SIMILARITY MEASURES OF PYTHAGOREAN FUZZY SETS


WITH APPLICATIONS TO PATTERN RECOGNITION AND
MULTICRITERIA DECISION MAKING WITH
PYTHAGOREAN TOPSIS
Zahid Hussain1, Sahar Abbas2, Shahid Hussain3, Zaigham Ali4
Gul Jabeen5
1,2,3
Department of Mathematical Science, Karakoram International University,
Gilgit-Baltistan, Pakistan
4
Department of Business Management, Karakoram International University,
Gilgit-Baltistan, Pakistan
5
Department of Computer Science, Karakoram International University,
Gilgit-Baltistan, Pakistan
zahid.hussain@kiu.edu.pk, taiys14125@gmail.com, zaigham.ali@kiu.edu.pk,
gul.jabeen@kiu.edu.pk, shahid8310@yahoo.com
Corresponding Author : Zahid Hussain

https://doi.org/10.26782/jmcms.2021.06.00006
Abstract
The construction of divergence measures between two Pythagorean fuzzy sets
(PFSs) is significant as it has a variety of applications in different areas such as
multicriteria decision making, pattern recognition and image processing. The main
purpose of this study to introduce an information-theoretic divergence so-called
Pythagorean fuzzy Jensen-Rényi divergence (PFJRD) between two PFSs. The
strength and characterization of the proposed Jensen-Rényi divergence between
Pythagorean fuzzy sets lie in its practical applications which are very closed to real
life. The proposed divergence measure is utilized to induce some useful similarity
measures between PFSs. We apply them in pattern recognition, characterization of
the similarity between linguistic variables and in multiple criteria decision making.
To demonstrate the practical utility and applicability, we present some numerical
examples related to daily life with the construction of Pythagorean fuzzy TOPSIS
(Techniques of preference similar to ideal solution). Which is utilized to rank the Belt
and Road initiative (BRI) projects. Our numerical simulation results show that the
suggested measures are well suitable in pattern recognition, characterization of
linguistic variables and multi-criteria decision-making environment.
Keywords: Divergence measure, Intuitionistic fuzzy set (IFS), Pythagorean fuzzy set
(PFS), Pattern recognition, similarity measure, multicriteria decision making.
I. Introduction
The idea of fuzziness initially comes from the concept of multi-valued logic.
Fuzzy set theory familiarized by Zadeh [XLVI] in 1965that allows the slow and
moderate valuation of the membership of the element belong to a set described by the
Zahid Hussain et al

64
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86
support of a real valued membership function in the unit interval (0,1). A declaration
like “The weather is hot” does not have a binary truth value rather it has fuzzy or
vague truth values between 0 and 1. The concept of hot has no clean boundary and it
varies from person to person. In reality, there is much fuzzy information that is
uncertain, vague, imprecise and probabilistic in nature. Fuzzy sets (FSs) have been
able to provide solutions to various actual-world problems. The term fuzziness or
uncertainty emerges when the piece of information is not clearly defined. There are
several methods to model uncertainty with high precision, for instance, distance and
similarity between two objects and the divergence which measures the discrepancy
between two objects. In daily life settings, we normally compare the illustration of
two objects and hence the process of comparison is very important. In literature,
several methods are being utilized to measure the difference between two objects
which are useful for many practical applications. For example, He et al. [XIV] in
2003 suggested the entropy based on generalized divergence with a new approach to
image registration. Lin [XVI] in 1991 proposed “a novel class of information-
theoretic divergence measures based on Shannon entropy”. The concept got enough
popularity in the context of fuzzy sets. Montes et al. [XIX] in 2002 proposed
divergence as a way of measuring the difference between two fuzzy sets using a
function. Bouchon et al. [VIII] in 1996 suggested classification of measures allowing
to associate fuzzy characterization of objects rendering to their properties and the
resolution of their application. Among the other generalization of fuzzy sets for
various objectives, the novel and powerful approach coined by Atanassov [VII]
in1986, Atanassov [V] in 1989, Atanassov [IV] in 1994a and Atanassov [II] in 2000,
an intuitionistic fuzzy sets (IFSs) which are more capable and useful than FSs.
Further, it extended to interval-valued IFSs by Atanassov and Gargov [VI] in 1989.
IFSs represents the degree of membership, the degree of nonmembership and the
degree of indeterminacy. They are more efficient to model the conditions in which
fuzzy sets do not afford enough evidence as in a variety of areas such as new
similarity measures of intuitionistic fuzzy sets centered on the Jaccard index with its
applications to clustering Ming et al. in [XVII] in 2018. Belief and plausibility
functions on intuitionistic fuzzy sets Hwang and Yang [13]in 2016. Intuitionistic
fuzzy cognitive maps Papageorgiou and Iakovidis [XII] in 2013, Analysis of
similarity measures for Atanssov’s intuitionistic fuzzy sets Szmidt and Kacprzyk
[XVIII] in 2009, Distances between intuitionistic fuzzy sets Szmidt and Kacprzyk
[XXX] in 2000, Preference relations based on intuitionistic multiplicative information
Xia et al. [XXXV] in 2013, Intuitionistic fuzzy compactness Abbas [I] in 2005.
Recently another generalization of FS sets called Pythagorean fuzzy sets (PFSs) has
been presented by the Yager [XXXIX] in 2013, Yager and Abbasov [XXXIX] in
2013 and Yager [XXXVII] in2014. PFSs are described by the square of membership
degree, non-membership degree and degree of uncertainty, an effective tool to model
such situations where intuitionistic fuzzy sets unable to consider all obtainable
evidence in decision making Yang and Hussain [XXXVII] in 2018, Multiparametric
similarity measures on Pythagorean fuzzy sets with applications to pattern
recognitionPeng and Garg [XX] in 2019and many authors published papers based on
distance, similarity measures and divergence of Pythagorean fuzzy sets i.e. Hussain
and yang [XII] in 2019, Garg [X], Peng and Yang [XXI], Ren et al. [XVI], Gou et al.
Zahid Hussain et al

65
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86
[XI] in 2016 and Xiao and Ding [XXXIV] in 2019. A novel divergence measure of
Pythagorean fuzzy sets, which is centered on the belief function in Dempster–Shafer
evidence theory and its applications in medical diagnosis Zhou et al. [XLII]in 2020.
Divergence is used in measuring the difference between two FSs. As far as we, know
there is no paper on Pythagorean fuzzy Jensen-Rényi divergence and its applications
to pattern recognition and multi-criteria decision making.
We suggest an axiomatic definition of a divergence measure for Pythagorean fuzzy
sets. Based on the definition a Pythagorean fuzzy Jensen-Rényi divergence and its
applications to pattern recognition and multi-criteria decision making are proposed.
Since fuzzy sets were based on the single membership value between 0 and 1. The
non-membership degree is not always (1-membership) because sometimes the
hesitancy is present. Then, to develop more determined consistency and applicability
Atanassov [III] in 1999generalized the idea of fuzzy set theory and suggested IFSs,
which contain both membership degree and non-membership degree and degree of
non-determinacy (degree of non-determinacy = (1- (degree of membership + non-
membership degree)), satisfying the condition that
 + 1 (1)
Further, Atanassov [3] in 1999 mentioned that there are IFSs over different universes
and IFSs of type 2 for which (1) is changed with the following:
2 +  2  1 (2)
We can see that (2) is the extension of (1) which is a normal fuzzy set but according
to Atanassov [III], such sets are not valuable and applicable. Now, Yager [39] in 2013
and Yager [XXXVIII] in 2014 recently, complete a wonderful and stunning change
on (2) and presentedPFSs fulfilling the required circumstance  2 +  2  1 with
different accumulation procedures and applications in multi-criterion decision
making. The condition when the figures  = 0.8660 and  = 0.5 we know since
 +  = 1.3660  1  2 = ( 0.86603)
2
and = 0.7500079609 = 0.75 and
 2 = ( 0.5) = 0.25 then  2 +  2 = 1. Firstly, PFSs are healthier than IFSs in
2

undertaking daily life hitches carrying vagueness and ambiguity. In this research, we
work and suggest information-theoretic divergence measure so-called Pythagorean
fuzzy Jensen-Rényi divergence measure with applications to pattern recognition and
multicriteria decision making.
The paper is organized as follows: In section 2, we briefly introduce some well-
known notions of IFSs and PFSs respectively and fix some notations and definitions.
In section 3, we introduce an axiomatic definition of the notion of divergence for
PFSs. Based on this definition, we establish a novel Pythagorean fuzzyJensen-Rényi
divergence between two Pythagorean fuzzy sets. We then induce some new and
useful similarity measures from the proposed PFJRD. In section 4, we utilize
proposed similarity measures in pattern recognition and develop a characterization of
the similarity between linguistic variables. Finally, we construct Pythagorean fuzzy
TOPSIS by using suggested similarity measures and exploit them in an application to

Zahid Hussain et al

66
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86
rank the Belt and road initiative (BRI) projects.In section 5, we conclude our
discussion.
II. Preliminaries
II.i. Intuitionistic and Pythagorean Fuzzy sets
We review basic definitions of intuitionistic fuzzy sets (IFSs) and Pythagorean
fuzzy sets (PFSs) to clearly differentiate between IFSs and PFSs.
Definition 1. Let us suppose that an intuitionistic fuzzy set (IFS) M in B is defined
by Atanassov [VII] in 1986 and Atanassov [III] in 1999as the following form:

M = ( r , 
( ri ) , M ( ri ) ) : ri  B ,
i M
(3)

where the functions M ( ri ) : B → 0,1 , represent the membership degree of ri  M


and  M ( ri ) : B → 0,1 , represent the non-membership degree of ri  M . For every
ri  B with the state that 0  M ( ri ) + M ( ri )  1. The degree of uncertainty or (non
- determinacy) of the element ri  B to IFS M is represented by the following
(
relation  M ( ri ) = 1 − M ( ri ) + M ( ri ) . )
Definition 2 A Pythagorean fuzzy set PFS G in B is given by Yager [XXXX] in
2013 and Yager [XXXVIII] in 2014as

G=  r, i G ( ri ) , G ( ri ) 
: ri  B (4)

where the functions G ( ri ) : B →  0,1 , represent the membership degree of ri  G

and  G ( ri ) : B →  0,1 , represent the non-membership degree of ri  G . For every


ri  B with the state that

0  G2 ( ri ) +  G2 ( ri )  1.

( )
Definition 3Atanassov [3] (1999). For all, G  PFSs B , we call the Pythagorean

index of the element ri  G , as the following expression,

 
 G ( ri ) = 1 − G2 ( ri ) + G2 ( ri ) , it is clear that 0   G2 ( ri )  1,  ri  B.

It is well-meaning to note that for PFS G ,


if G2 ( ri ) = 0.0, then  G2 ( ri ) +  G2 ( ri ) = 1.0, and if G2 ( ri ) = 1.0 then
 G2 ( ri ) = 0.0 and  G2 ( ri ) = 0.0. Similarly  G2 ( ri ) = 0.0, then
G2 ( ri ) +  G2 ( ri ) = 1.0, and if  G2 ( ri ) = 1.0 then G2 ( ri ) = 0.0 and  G2 ( ri ) = 0.0,

Zahid Hussain et al

67
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86

and if  G2 ( ri ) = 0.0 then G2 ( ri ) + G2 ( ri ) = 1.0, and if  G2 ( ri ) = 1.0 ,


G2 ( ri ) =  G2 ( ri ) = 0.0.
Definition 4(Yang, Hussain, 2018) [XXXVI] We will say that a Pythagorean fuzzy
set is completely Pythagorean if G2 ( ri ) =  G2 ( ri ) = 0.0,  ri  B.
Definition 5 (Yang, Hussain, 2018)[36]If G and Q are two PFSs of the set B , then
(i) G  Q if and only if  ri  B, G2 ( ri )  Q2 ( ri ) and  G ( ri )   Q ( ri ) ;
2 2

(ii) G = Q if and only if  ri  B, G ( ri ) = Q ( ri ) and  G ( ri ) =  Q ( ri ) ;


2 2 2 2

(iii) G =
c
 r ,
i G ( ri ) , G ( ri ) 
: ri  B .

II.ii. Jensen-Rényi Divergence


In this subsection, we briefly discuss the context of probability. Suppose that for all
n  N , B = r1 , r2 ,..., rn  be a finite set with a probability distribution
P =  p1 , p2 ,..., pn  , then we can represent the set of n-complete ( n  2 ) probability
distributions as follows
 n

Bn =  P =  p1 , p2 ,..., pn  :  pi = 1, and pi  0  , P  Bn .
 i =1 
The well-known Shannon’s entropy Zhang et al. [43]can be defined as
n
H ( P ) = − pi log pi
i =1
It is worthy to mention that the logarithm base 2 will be used because the
measurement unit is bits and also for the sake of computation easiness. Rényi
generalized the notion of Shannon’s entropy and introduced the Rényi’s entropy by
Rényi [XXVIII] of order  , which can be stated as
1  n 
R ( P ) = log   pi  ,   1,   0.
1−  i =1 
The Rényi entropy is neither concave nor convex for   1. It is straightforward to
perceive that R ( P ) is concave for   ( 0,1) , and as  → 1, Rényi entropy
R ( P ) tends to Shannon’s entropy H ( P ) . Further, it can be justified that R P ( )
is a non-increasing function for   ( 0,1) , consequently

( ) ( )
H P  R P ,   ( 0,1) .
In continuation,  is confined to the open interval (0,1) i.e   ( 0,1) , unless
otherwise specified. More specifically, we first use Lin [XVI] to define the Jensen-
Shannon divergence measure for two probability distributions with weights. Let

Zahid Hussain et al

68
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86

w1 , w2  0, such that w1 + w2 = 1, be the arbitrary weights of two probability


distribution P1 , P2  Bn respectively, then the Jensen-Shannon divergence can be
defined as
( ) (
JSw P1 , P2 = H w1P1 , w2 P2 − w1H P1 − w2 H P2 ) ( ) ( ) (5)

( )
where H P is Shannon’s entropy and is a concave function, as stated by Jensen’s

inequality that JS w P1 , P2 ( ) is non-negative and when P1 = P2 it disappears


completely. The main feature of the Jensen-Shannon divergence which is particularly
useful in decision-making problems is that we can allocate any weights to the
probability distribution according to their priority.
Suppose that P1 , P2 ,..., Pn be n finite probability distributions on B , and further,
assume that weight of every element ri  B is wi ( i =1, 2,3,..., n ) such that
n

w
i =1
i = 1 , where 0  wi  1. Then the Jensen-Rényi divergence suggested by HE et

al. (2003) as
 n  n
( )
JRw, P1 , P2 ,..., Pn = R   wi Pi  −  wi R Pi ( ) (6)
 i =1  i =1
where R P ( ) is the Rényi entropy with   0 and   1. Considering two
probability distributions (6) can be rewritten as
( ) ( )
JRw, P1 , P2 = R w1P1 + w2 P2 − w1R P1 − w2 R P2 , ( ) ( )   ( 0,1) . (7)
The (7) is the generalized form of (5) that shows Jensen- Rényi divergence for two
probability distributions.
Shannon [XXXI] first introduced in communication theory that divergence measures
are based on the idea of information-theoretic entropy. The most widely used method
of statistical implication is divergence measure founded on entropy functions which
denote the distinction between two probability distributions.
III. Pythagorean Fuzzy Jensen-Rényi Divergence (PFJRD)
Let a PFS P is defined on a singleton element universal set B = ri . For the
sake of simplicity, PFS P is denoted by triplet ( P , P ,  P ) such that
 P2 +  P2 +  P2 = 1, 0  P2 , P2 ,  P2  1, then ( P , P ,  P ) may be considered as a
probability distribution. Therefore, we may use the above concept to measure the
difference between two PFSs. Let us assume that P and Q respectively are two
Pythagorean fuzzy sets defined on a universal set composed of a singleton element
(
B = ri  . P = ri ,  P ( ri ) , P ( ri ) ,  P ( ri ) : ri  B ) with the condition that

P2 ( ri ) + P2 ( ri ) +  P2 ( ri ) = 1 , where the functions P ( ri ) : B → 0,1 represent the


Zahid Hussain et al

69
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86

membership degree of ri  P and the function  P ( ri ) : B → 0,1 represents the

(
degree of non-membership of ri  P . Similarly, Q = ri , Q ( ri ) , Q ( ri ) ,  Q ( ri ) : ri  B )
with the condition that Q2 ( ri ) + Q2 ( ri ) +  Q2 ( ri ) = 1 where the function
Q ( ri ) : B →  0,1 represents the membership degree of ri  Q the functions
 P ( ri ) : B → 0,1 and  Q ( ri ) : B →  0,1 represents the degree of nonmembership
of ri  P and ri  Q respectively. For every ri  B ,  P2 ( ri ) = 1 −  P2 ( ri ) − P2 ( ri )

and  Q2 ( ri ) = 1 − Q2 ( ri ) − Q2 ( ri ) are called the Pythagorean fuzzy index of the

element ri  B to the PFSs P and Q respectively.


Definition 6Let P , Q and R be three PFSs on B . Then D is called divergence
between PFSs if it satisfies the conditions ( C1 − C5 )
( C1 ) ( )
0  D P, Q  1;

( C ) D ( P, Q ) = 0, iff P = Q ;
2

(C ) D ( P, Q ) = D ( Q, P )
3

(
(C4 ) P  Q  R, then D( P, R)  max D( P, Q), D(Q, R) ; )
(C5 ) For P and Q, then D( P  Q, P  Q) = D( P, Q) and D( P  Q, P  Q) = D(Q, P).

( ) (
Now, we consider P ( ri ) , P ( ri ) ,  P ( ri ) and Q ( ri ) , Q ( ri ) ,  Q ( ri ) two )
probability distributions analogous to (11), we can define Pythagorean fuzzy Jensen-
Rényi divergence measure between two PFSs P and Q as

( ) (
JRDPFS w, P, Q = R w1 P + w2Q − w1R P − w2 R Q ) ( ) ( )
where   ( 0,1) and w1 + w2 = 1, w1 , w2  0, and

 ( ) ( ) (
w1P + w2Q = w1 ( P ( ri ) ) + w2 Q ( ri ) , w1 ( P ( ri ) ) + w2  Q ( ri ) , w1 ( P ( ri ) ) + w2  Q ( ri )
2 2 2 2 2
)
2

Zahid Hussain et al

70
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86
n  2  

JRDwPFS
,
(
P, Q = ) 1
   1 P i
log
n (1 −  ) i =1  
w (  ( r ) )
2
+ w2  (
Q i
r ) (
  1 P i
 
) ) 2(
 +w  r 2 + w  r 2 
( (
Q i ) 

) ( )
( ( ))  − w log ((  ) + (( ) + ( ( ) 
 2  2  2 
+ w1 ( P ( ri ) ) + w2  Q ( ri ) ( ri ) ) ( ri ) ) ( ri ) )
2
1 P P P

 2  2  2  
( ) (
− w2 log  Q ( ri )  +   Q ( ri )  +   Q ( ri )  
      
) ( ) (8)

where w1 , w2  0,1 represent the weights such that. Next, we briefly give the proof
of properties of (8) in the following theorems.
Theorem 1 Let B = ri  be a singleton universe of discourse. The proposed Jensen-
PFS
Rényi divergence JRDw, P, Q ( ( )) between two PFSs P and Q , satisfies all the
conditions of Definition 6.
Proof. To prove ( C1) Definition 6, we use the preposition 1 of He et al. [XLV] that
JRDwPFS
,
( )
P, Q is convex for   ( 0,1) . Therefore, for   ( 0,1) , JRDwPFS
,
P, Q ( )
increases as P − Q increases, where
1

P − Q =  P2 ( ri ) − Q2 ( ri ) +  P2 ( ri ) − Q2 ( ri ) +  P2 ( ri ) −  Q2 ( ri )
1

Hence,    ( 0,1) , JRDw ,


PFS
( P, Q ) attains its maximum for the following degenerate
cases
P = (1.0, 0.0, 0.0 ) , Q = ( 0.0,1.0, 0.0 ) or P = ( 0.0,1.0, 0.0 ) , Q = ( 0.0, 0.0,1.0 ) or
P = ( 0.0, 0.0,1.0 ) , Q = (1.0, 0.0, 0.0 ) which gives 0  JRDwPFS
,
P, Q  1 . Proving ( )
( C 2 ) of Definition 6, the result directly from the Jensen’s inequality. Now, we prove
( C 3) Definition 6,
1   2  2 
JRDwPFS
,
(
P, Q = )  
( ) 
log  w1 (  P ( ri ) ) + w2 Q ( ri )  +  w1 ( P ( ri ) ) + w2  Q ( ri ) 
(1 −  )  
2

 
2


( )
( ( )) ( ) + (( ) + ( ( ) 
  2  2 
 
+ w1 ( P ( ri ) ) + w2  Q ( ri )  − w1 log  (  P ( ri ) ) ( ri ) ) ( ri ) )
2 2

 
P P

 2  2  2  
( ) ( )
− w1 log  Q ( ri )  +   Q ( ri )  +   Q ( ri )  
      
( )
1   2 
(( ) )
 

( ) 2 
( ) ( ( ))   ( ) ( ( )) 
2
( )
( ) ( ( )) 
2 2
=  
log w  r + w  r + w  r + w  r + w  r + w  r
(1 −  )   2 Q i 1 P i   2 Q i 1 P i  2 Q i 1 P i 

( ) ( ) ( )
 2  2  2  


( ) ( ) ( )

 
 
 

 


2  2  2 
− w1 log  Q ( ri )  +   Q ( ri )  +   Q ( ri )   − w1 log  ( P ( ri ) ) + ( P ( ri ) ) + ( P ( ri ) ) 

Zahid Hussain et al

71
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86

= JRDwPFS
,
( )
Q, P . Now, we give the proof of ( C 4 ) of Definition 6, for P, Q , R  X ,

if PQR P −Q  P − R and Q−R  P−R .Thus,


1 1 1 1

( ) ( )
JRDPFS w, P, Q  JRDPFS w, P, R and

JRDPFS ( Q, R )  JRDPFS ( P, R ) ,
w , w ,

hence JRDPFS ( P, R )  max ( JRDPFS ( P, Q ), JRDPFS


w, w, w, )
(Q, R ) .
Finally, we give the proof of ( C 5 ) of Definition 6, from (8), we have

JRDwPFS
,
(
P  Q, P  Q = ) 1 
(1 −  ) 
log w(
1  2
P  Q ( ri ) + w2  2
P  Q ( ri )

+ w1 2
P) (
 Q ( ri ) + w2 2
P  Q ( ri )

)
( ( ) ( ))  − w log (  ( ri )) ( )
  
+ w1 1 −  P2Q ( ri ) − P2Q ( ri ) + w2 1 −  P2Q ( ri ) − P2Q ( ri ) 1
2
P Q
+  P2Q ( ri )

( )  − w log (  ( ri )) ( ) ( ) 
   
+ 1 −  P2Q ( ri ) − P2Q ( ri ) 2
2
P Q
+  P2Q ( ri ) + 1 − P2Q ( ri ) − P2Q ( ri )

=
1 

(1 −  ) 
log w1  2
Q ( (
ri ) + w2  2
P ( ri )

+ w1 2
Q ( ri ) + w2) (
 2
P ( ri )

+ w1 1 − Q2 ( ri ) − Q2 ( ri ) ) ( ( )
( ) )  − w log (  ( r )) + ( ( r )) + (1 −  ( r ) − ( r )) 
   
+ w2 1 −  P2 ( ri ) − P2 ( ri ) 1
2
Q i
2
Q i
2
Q i
2
Q i

( ) ( ) ( )  + log ( w  ( r ) + w  ( r ))
   
− w2 log  P2 ( ri ) +  P2 ( ri ) + 1 −  P2 ( ri ) − P2 ( ri ) 2
1 P i 2
2
Q i

( ) ( ( ) ( )) 
 
+ w1 P2 ( ri ) + w2 Q2 ( ri ) + w1 1 −  P2 ( ri ) − P2 ( ri ) + w2 1 −  P2 ( ri ) − P2 ( ri )

( ) ( ) ( )  − w log (  ( r )) + ( ( r ) )
    
− w1 log  P2 ( ri ) +  P2 ( ri ) + 1 −  P2 ( ri ) − P2 ( ri ) 2
2
Q i
2
Q i

( )  = JRD ( P, Q ).

+ 1 − Q2 ( ri ) − Q2 ( ri ) PFS
w ,

IV. Application to Pattern recognition and multicriteria decision making with


TOPSIS
It is worthy to mention that the Jensen- Rényi divergence presents the
dissimilarity between two PFSs and therefore it can be used for a variety of
applications such as pattern recognition, multi-criteria decision making (MCDM),
clustering etc. Instead of dissimilarity measures between two PFSs, it may be
accepted and more useful to reflect a similarity measure between two PFSs. Then the
similarity measure may be demarcated by an appropriate decreasing function of
divergence measured, such as 1-d, specifically the range of d is (0,1). The similarity
Zahid Hussain et al

72
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86

measure between PFSs P and Q in X based on the proposed Jensen- Rényi


divergence JRD wPFS
,
( )
P, Q can be formulated as follows.

Let the monotone decreasing function be g . Since 0  JRDwPFS


,
P, Q  1 then ( )
(
 (1)   JRD PFS
 w,
)
( P, Q )   ( 0 ) provided that  (1)   ( 0) ,
Which implies

0
 JRDwPFS
,
( P, Q −  (1)
 1.
( ))
 ( 0 ) −  (1)
Therefore, the similarity measure between PFSs P and Q is given by
(
 JRD PFS ( P, Q ) −  (1) )
(
S P, Q = )  w,

 ( 0 ) −  (1)
(9)

( )
Constructed on theorem 1, S P, Q fulfills the following axioms ( A1 − A5 ) .

( )
Definition 7. A function S : PFSs B  PFSs B →  0,1 . S P, Q is said to be ( ) ( )
the degree of similarity between P  PFSs ( B ) and Q  PFSs ( B ) if S ( P, Q )
satisfies the axioms ( A1 − A5 ) :
( A1) ( )
0  S P, Q  1;

( A2 ) S ( P, Q ) = 1, if P = Q ;

( A3) S ( P, Q ) = S ( Q, P )

( A4) If P  Q  R  PFSs ( X ) then S ( P, R )  min S ( P, Q ), S (Q, R ) ; ( )


( A5) For P and Q, then S ( P  Q, P  Q ) = S ( P, Q ) and S ( P  Q, P  Q ) = S (Q, P ).

Remark For applications related to pattern recognition, Mitchell [XVIII] proposed


replacing ( A2 ) with its stronger version ( A2 ) as follows

( A2) ( )
S P, Q = 1, iff P = Q.
The main issue here is how to choose the simplest and appropriate function  . The
most suitable function  can be chosen as  = 1 − x . Then the similarity function
based on the Pythagorean fuzzy divergence between two PFSs P and Q using (9) is
denoted as follows
( )
S md P, Q = 1 − JRD wPFS
,
P, Q ( ) (10)
Next, we may select a simple and suitable rational function as follows
1
 =
1+ x

Zahid Hussain et al

73
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86
Then the similarity function based on Pythagorean fuzzy divergence between two
PFSs P and Q using (9) is denoted as follows
1 − JRDwPFS P, Q ( )
(
S nd P, Q = ) ,
(11)
1 + JRDw , PFS
( P, Q )
Next, we choose an exponential function which is extremely beneficial in allocating
with similarity relation in a variety of applications such as Zadeh [XLV] in 1971,
entropy a novel definition and its applications Pal and Pal [25] in 1991, some
properties of the exponential entropy Pal and Pal [XXIV] in 1992 and clustering
analysis Wu and Yang [XXXIII] in 2002 and Yang and Wu [XLI] in 2004. The chose
exponential can be defined as
 = exp ( − x )
Then the similarity function based on the Pythagorean fuzzy divergence between two
PFSs P and Q using (9) is denoted as follows
− JRDwPFS ( )
− e−1
P ,Q

( )
,
e
S zd P, Q = (12)
1 − e−1
IV.i. Pattern recognition
In this subsection, we give some examples to demonstrate the unsuitability of our
suggested similarity measures in the environment of pattern recognition. In this paper,
we consider the weights w1 = 0.7 and w2 = 0.3 in all numerical examples.
Example 1 Assume that there are three patterns represented with PFSs in
B = 30, 40,50. These three patterns are denoted as follows
P1 = ( 30, 0.7, 0.7 ) , ( 40, 0.5, 0.5) , ( 50, 0.3, 0.3) ;
P2 = ( 30,0.5,0.5) , ( 40,0.5,0.5) , (50,0.5,0.5);
P3 = ( 30,0.3,0.3) , ( 40,0.3,0.3) , (50,0.3,0.3).
Assume that a sample Q = (30,0.7,0.7 ) , ( 40,0.5,0.5) , (50,0.3,0.3) is given.
( ) ( ) ( )
S md P1 , Q = 1, S md P2 , Q = 0.9412, S md P3 , Q = 0.8839;

S nd ( P , Q ) = 1, S ( P , Q ) = 0.8889, S ( P , Q ) = 0.7920;
1 nd 2 nd 3

S zd ( P , Q ) = 1, S ( P , Q ) = 0.9097, S ( P , Q ) = 0.8266.
1 zd 2 zd 3

From the above data, it can be clearly seen that P1 = Q that is, the sample Q belongs
to the pattern P1 . It shows if two PFSs are exactly the same then the proposed
similarity (
measures S md P1 , Q , S nd P1 , Q ) ( ) and (
S zd P1 , Q ) between two
PFSsdisplay the precise ordering rendering to the norm of maximum membership
degree. As we take PFS P2 a little away from the sample Q then the divergence

Zahid Hussain et al

74
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86

between P2 and the sample Q get increased. In other words, the more similar two
PFSs are, the lower the divergence between them. Our proposed similarity measures
reflect the precise ordering rendering to the norm of maximum similarity. A similar
result can be seen between the pattern P3 sample Q which is analogous to our
intuition.
Example 2 Let us consider that B = 30, 40,50 are three patterns denoted with
PFSs. These three patterns are denoted as follows

P1 = ( 30, 0.6, 0.6, 0.5292 ) , ( 40, 0.7, 0.5, 0.5099 ) , (50, 0.2, 0.3, 0.9327 ) ;
P2 = ( 30, 0.5, 0.5, 0.7071) , ( 40, 0.3, 0.9, 0.3162 ) , ( 50, 0.1, 0.4, 0.9110 ) ;
P3 = ( 30,0.7,0.7,0.1414 ) , ( 40,0.6,0.6,0.5292 ) , (50,0.3,0.2,0.9327 );

Let th ae sample

Q = ( 30,0.7,0.7,0.1414 ) , ( 40,0.5,0.7,0.5099 ) , (50,0.0,0.5,0.8660 ) is given.

( ) ( ) ( )
S md P1 , Q = 0.9415, S md P2 , Q = 0.9278, S md P3 , Q = 0.9497;

S nd ( P , Q ) = 0.8895, S ( P , Q ) = 0.8653, S ( P , Q ) = 0.9042;


1 nd 2 nd 3

S zd ( P , Q ) = 0.9101, S ( P , Q ) = 0.8898, S ( P , Q ) = 0.9224.


1 zd 2 zd 3

From the calculations of example 2, we infer that the sample Q belongs to the patter
P3 according to the norm of maximum similarity which is reflected by our proposed
similarity measures. Hence, we can say that our suggested measures are appropriate
and well suitable in the recognition of patterns.
IV.ii. Classification of the similarity between linguistic variables
We have demonstrated the consistency and applicability of our suggested similarity
measures. Furthermore, we claim that our proposed similarity measures are also
reasonable in the representation of the similarity between linguistic variables. To
support our claim, we exhibit an example related to the representation of the
similarity between linguistic variables by using our suggested similarity measures
Eqs. (10) - (12). In the recent past, the fuzzy database studied widely cf. Petry
[XXIII] in 1996. Similarity measures play a vital role in querying a database cf.
Candan et al. [IX] in 2000. The similarity between linguistic variables is
characterized by our suggested similarity measure between PFSs as follows:
Example 3. Assume that a PFS on a universal set B is given by
P=  r,
i P ( ri ) , P ( ri ) : ri  B
n
The nth power for any positive number n, we have the PFS P Yang and Hussain
[XXV] with

Zahid Hussain et al

75
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86

 
(
P n =  ri , (  P ( ri ) ) , 1 − 1 − P2 ( ri ) )
n n
: ri  B  .
 
In addition, we consider two linguistic operators, CON ( P ) = P 2 , represent the
1
Concentration of P : DIL ( P ) = P 2 represent and Dilation of P . Now, we assume
PFSs P in B = 10, 20,30, 40,50 as

P =  10, 0.5, 0.7 , 20, 0.4, 0.8 , 30, 0.3, 0.9 , 40, 0.2, 0.3 , 50, 0.0,1.0 . Thus,
we have
1
P 2 regarded “ less LARGE”,
P2 regarded “Very LARGE”,
P 5 regarded “Very very LARGE”.
P =  1,0.5,0.7,0.5099 , 2,0.4,0.8,0.4472 , 3,0.3,0.9, 0.3162 , 4,0.2,0.3,0.9327 , 5,0.0,1.0,0.0 .
1
P 2 =  1,0.7071,0.5347,0.4627 , 2,0.6325,0.6325,0.4471 , 3,0.5477,0.7511, 0.3686 ,
4,0.4472,0.2146,0.8683 , 5,0.0,1.0,0.0 ;
P 2 =  1,0.2500,0.8602,0.4445 , 2,0.1600,0.9330,0.3224 , 3,0.0900,0.9818,0.1672 ,
4, 0.0400,0.4146,0.9091 , 5,0.0,1.0,0.0  ;
P5 =  1,0.0313,0.9826,0.1831 , 2,0.0102, 0.9970, 0.0767 , 3,0.0024,0.9999,0.0139 ,
4, 0.0003,0.6132,0.7899 , 5,0.0,1.0,0.0  ;
We use acronyms L. for LARGE, M.L.L. for More or less Large, V.L. for Very
Large, and V.V.L. for Very very LARGE. To calculate the degrees of similarities
between PFSs the suggested similarity measures Eqs. [X-XII] are used.Thus, we
acquire the following preconditions rendering to degrees of similarities between PFSs
from Table 1:
S ( L., M.L.L.)  S ( L., V.L.)  S ( L., V.V.L.) ,
S ( M.L.L., L.)  S ( M.L.L., V.L.)  S ( M.L.L., V.V.L.) ,

S ( V.L., V.V.L.)  S ( V.L., L )  S ( V.L., M.L.L.) ,


S ( V.V.L., V.L )  S ( V.V.L., L.)  S ( V.V.L., M.L.L.).

Zahid Hussain et al

76
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86
Table 1. Degrees of similarities using Eqs. [X-XII]
L. M.L.L. V.L. V.V.L.
L. 1.0000(12) 0.9717(12) 0.9603(12) 0.8473(12)
1.0000(13) 0.9450(13) 0.9236(13) 0.7351(13)
1.0000(14) 0.9559(14) 0.9384(14) 0.7760(14)
M.L.L. 0.9717(12) 1.0000(12) 0.9002(12) 0.7768(12)
0.9450(13) 1.0000(13) 0.8185(13) 0.6351(13)
0.9559(14) 1.0000(14) 0.8497(14) 0.6835(14)
V.L. 0.9603(12) 0.9002(12) 1.0000(12) 0.9311(12)
0.9236(13) 0.8185(13) 1.0000(13) 0.8711(13)
0.9384(14) 0.8497(14) 1.0000(14) 0.8947(14)
V.V.L. 0.8473(12) 0.7768(12) 0.9311(12) 1.0000(12)
0.7351(13) 0.7351(13) 0.8711(13) 1.0000(13)
0.7760(14) 0.7760(14) 0.8947(14) 1.0000(14)

IV.iii. Application to multicriteria decisionmaking with Pythagorean fuzzy


TOPISIS
Decision-making is a process through which the decision-maker reached a
certain decision. There are many methods of PFS decision-making like TODIM,
VIKOR etc. In this study, we use TOPSIS for multicriteria decision-making and
apply it with the help of particle examples. The decision needs to be wise, reasonable,
purposeful and objective oriented and swift. It is a kind of daily activity where one
needs to exercise it a moment to moment to make the activity useful. During the
process of reasonable decision-making, individuals will analyze alternatives and
extract consequences from each alternative and finally choose the best out of it.
Decision-making plays a vital role in many areas such as business management,
social and political sciences etc.
Multi criterion decision making (MCDM) is a method to choose the best
alternative from a finite number of possible and suitable alternatives under the
umbrella of many criteria. To acquire the best alternative, decision-makers deliver
their predilection evidence among a set of available alternatives. The information
available in the actual life decision-making procedure is frequently fuzzy, ambiguous
and imprecise in nature. PFSs are initiated to be the best implement to resolve
decision-making problems containing fuzzy, unclear, or inaccurate evidence with
high accuracy. In this context, several multicriteria decision-making processes
centered on PFSs were found to be more actual and precise. In the following section,
we apply the suggested similarity measures Eqs. [X-XII] based on the Jensen- Rényi
divergence to MCDM problems. It assists in the choice of the best or most preferred
Zahid Hussain et al

77
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86
alternative among an existing set of alternatives. We use the idea of TOPSIS to
paradigm an algorithm for Pythagorean fuzzy TOPSIS based on similarity measures
to solve MCDM problems.
The steps for Pythagorean fuzzy TOPSIS based on the suggested similarity
measures are as follows:
Step 1: Construction of Pythagorean Fuzzy decision matrix (PFDM)
We represent estimate values of alternatives G = G1 , G2 ,..., Gm  , with

( ) (  ) 
reverence to the criteria C = C1 , C2 ,..., Cn , by Gi C j = P ij , ij ,  ij . Then,

D = ( G ( C ) ) denotes the PFDM. Here, P (  , ,  ) symbolizes Pythagorean


i j ij ij ij
m n
fuzzy number (PFN) that represents estimate ideals specified by decision-makers so
that the alternative Gi placates the criteria C j . The PFDM is represented as follows;
C1 . . . Cn
G1  r11 . . . r1n 
. 
 . . . 
D = (G (C )) i j
m n
= .  .

. . 

.  . . . 
Gm 
 rm1 . . . rmn 

Step 2: Pythagorean fuzzy positive ideal solution (PFPIS) and Pythagorean


fuzzy negative ideal solution (PFNIS)
The valuation criteria can be classified into two groups, i.e. benefit and costs in
TOPSIS method. Let us considered the collection of benefit criteria be M and the
− +
collection of cost criteria be N . Thus, we define the PFNIS I and PFPIS I as
follows:

 ( ( ) ) ( ( ) ) ( ( ) ) ( ( ) ) : i  m ,
I + =  C j , max ij C j : j  M , min ij C j : j  N , min ij C j : j  M , min ij C j : j  N
 i i i i

I = C , ( min  ( C ) : j  M ) , ( max  ( C ) : j  N ) , ( max ( C ) : j  M ) , ( min (C ) : j  N )  : i  m  .



− 
 
j ij j ij j ij j ij j
i i i i

Step 3: Construction of similarity measures between alternative Ai to PFPIS and


PFNIS
Now, we utilize our proposed similarity measures [X-XII] between PFSs P and Q
calculate the similarity measures of each alternative Gi from PFPIS and PFNIS
respectively.
S + ( Gi ) = JRDwPFS
,
Gi , I + , and ( )
S −
(Gi ) = JRD PFS
w ,
(G , I ) ,
i

i = 1, 2,..., m.

Step 4: Construction of degree of relative closeness


We can find the degree of relative closeness by the following expression:

Zahid Hussain et al

78
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86

S + ( Gi )
S ( Gi ) =
S + ( Gi ) + S − ( Gi )

Step 5: Ranking of the alternatives


The maximum value alternative will be measured as the most favored
alternative. We arrange the ranking of alternatives according to descending order.
Example 4. A case study to rank the plans under the Belt and Road initiative (BRI)
on a priority basis in the light of connected specialists’ opinion is employed to
validate the effectiveness of the multi-criteria decision-making method as proposed
earlier.
Belt and Road initiative (BRI) is a multibillion US dollar project mainly are
connected to infrastructure progress, energy, mining, IT and communication sectors,
industrial parks, tourist and urban development etc. It is a framework of regional
connectivity that connects Asia with Africa and Europe via land and maritime
networks along with six corridors to improve regional integration, people to people
contact, increasing trade and invigorating economic growth. It would be beneficial to
enhance the geographical linkages, improve the life standard of people, road, rail, air
transport system with regular and free arguments of growth and people to people
contact. (https://www.beltroad-initiative.com)
Assume that the related specialists are permissible to rank the BRI schemes on an
imports basis. Assume that, firstly there are five possible mega projects which are:
Central Asia-China Gas Pipeline ( G1 ) , Europe-China Rail Link ( G2 ) , Pakistan China–
Fiber Optic Project ( G3 ) Gilgit-KIU Hydropower ( G4 ) , and Forest City ( G5 ) under the

( )
following four criteria: Time frame for completion of the project C1 , Maintenance and

( )
sustainability C2 , Visible socioeconomic development C3 ( ) and Nature-friendly and
( )
green C4 . A detailed description of such criteria is displayed in following Table 2.
Table 2. Criteria for ranking BRI projects
Criterion Description
Time frame for Roadmap to ensure timely completion of project without any
completion of project C1 interruption and hurdle and also provide clear picture of the
work that needs to be done
Maintenance and The maintenance, repair, reliability and sustainability of the
sustainability C2 project

Visible Socio-economic To bring visible development and improvement in GDP,


development C3 economy stability and prosperity, life expectancy, education,
health, employment, personal dignity, personal safety and
freedom
Nature-friendly and green Those products that claims no harm upon ecosystem,
C4 contributes to green living, practices that help conserve natural
resources and prevent contribution to air, water and land
pollution

Zahid Hussain et al

79
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86
Suppose that a group of decision makers with five related field specialists is approved
to assess the defined alternatives concerning identified criterion. The intentions of the
alternatives Gi under the criteria C j are shown in Table 3, in the form of PFDM
D = ( pij ) , is given as follows
54

Table 3 Hesitance fuzzy decision matrix D


Q1 Q2 Q3 Q4
G1 0.3, 0.9, 0.3162 0.7, 0.7, 0.1414 0.6, 0.6, 0.5292 0.3,0.6,0.7416
G2 0.6, 0.6, 0.5292 0.9, 0.5, 0.3162 0.8, 0.4, 0.4472 0.9,0.3,0.3162
G3 0.4, 0.8, 0.4472 0.5, 0.7, 0.5099 0.3, 0.2, 0.9327 0.4,0.5,0.7681
G4 0.5,0.4,0.7681 0.3, 0.2, 0.9327 0.7, 0.7, 0.1414 0.1,0.4,0.9110
G5 0.7, 0.5, 0.5099 0.4,0.1,0.9110 0.1,0.8,0.5916 0.7,0.7,0.1414
+ −
The construction of PFPIS I and the PFNIS I by utilizing the data given in
Table 3 is displayed in the following Table 4
Table 4. Construction of PFPIS and PFNIS
I+ 0.7, 0.4, 0.5916 0.9, 0.1, 0.4243 0.8,0.2,0.5657 0.9,0.3,0.3162

I− 0.3,0.9,0.3162 0.3, 0.7, 0.6481 0.1, 0.8, 0.5916 0.1, 0.7, 0.7071

Next, we use our suggested similarity measures Eqs. [X-XII] to construct the degree
of similarity between each alternative Gi to PFPIS and PFNIS correspondingly. The
simulation outcomes are presented in the following Table 5.
Table 5. The similarity between each alternative to PFPIS and PFNIS respectively

S md S + ( Gi ) S − ( Gi ) Snd S + ( Gi ) S − ( Gi ) S zd S + ( Gi ) S − ( Gi )

G1 0.8492 0.9045 G1 0.7379 0.8257 G1 0.7786 0.8559


G2 0.9664 0.8173 G2 0.9350 0.6910 G2 0.9477 0.7358
G3 0.8616 0.9152 G3 0.7569 0.8437 G3 0.7955 0.8714
G4 0.7910 0.8480 G4 0.6543 0.7361 G4 0.7016 0.7769
G5 0.8496 0.8393 G5 0.7385 0.7231 G5 0.7791 0.7652

Table 5 is utilized to construct the relative closeness degree.

Zahid Hussain et al

80
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86

Table 6. Relative closeness coefficients for each alternative Gi


Smd S ( Gi ) Snd S ( Gi ) S zd S ( Gi )
G1 0.4842 G1 0.4719 G1 0.4764
G2 0.5418 G2 0.5750 G2 0.5629
G3 0.4849 G3 0.4729 G3 0.4772
G4 0.4826 G4 0.4706 G4 0.4745
G5 0.5030 G5 0.5053 G5 0.5045

The results of ranked alternatives are exhibited in Table 7.


Table 7. The ranking order of alternatives for different entropies
Similarity Ranking Best alternative
S md G2 G5 G3 G1 G4 G2
Snd G2 G5 G3 G1 G4 G2
S zd G2 G5 G3 G1 G4 G2

The ranking results in Table 7 show that our proposed similarity measures Eqs. [X-
XII] rank the alternative in similar order and there is no conflict in ranking the
alternative. Completely, numerical assessment outcomes and application clarify that
the suggested similarity measures Eqs. [X-XII] is much simpler and well suitable in a
multi-criterion decision-making environment.

Example 5. Suppose that a person wants to buy a mobile phone for his daily use.
There are four alternatives D1 , D2 , D3 and D4 (Mobile companies) with certain
criteria G1 , G2 , G3 and G4 and we consider these alternatives are four different
mobile companies. The criteria are as follows;

G1 : Price, G2 : Storage, G3 : Battery timing, G4 : Function.

D1 =  1 ,0.7,0.4 , 2 ,0.5,0.4 , 3 ,0.7,0.3 , 4 0.6,0.5 ;

D2 =  1 ,0.4,0.5 , 2 ,0.8,0.5 , 3 ,0.9,0.3 , 4 0.7,0.6 ;

D3 =  1 ,0.6,0.6 , 2 ,0.7,0.6 , 3 ,0.9,0.1 , 4 ,1.0,0.0 ;

D4 =  1 ,0.8,0.2 , 2 ,0.7,0.6 , 3 ,0.9,0.0 , 4 0.4,0.8 .

Zahid Hussain et al

81
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86
Table 8. Hesitance fuzzy decision matrix D

G1 G2 G3 G4
D1 0.7, 0.4, 0.35 0.5, 0.4, 0.59 0.7, 0.3, 0.42 0.6, 0.5, 0.39
D2 0.4, 0.5, 0.59 0.8,0.5,0.11 0.9, 0.3, 0.10 0.7, 0.6, 0.15
D3 0.6, 0.6, 0.28 0.7, 0.6, 0.15 0.9,0.1,0.18 1.0,0.0,0.00
D4 0.8, 0.2, 0.32 0.7, 0.6, 0.15 0.9, 0.0, 0.19 0.4, 0.8, 0.20
+ −
The construction of PFPIS I and the PFNIS I by utilizing the data given in
Table 8 is displayed in the following Table 9
Table 9. Construction of PFPIS and PFNIS
I+ 0.4,0.6,0.48 0.8, 0.4, 0.2 0.9, 0.1, 0.18 1.0, 0.0, 0.0
I− 0.8,0.2,0.32 0.5,0.6,0.39 0.7,0.3,0.42 0.4, 0.8, 0.2

Next, we use our suggested similarity measures Eqs. [X-XII] to construct the degree
of similarity between each alternative Gi to PFPIS and PFNIS correspondingly. The
simulation outcomes are presented in the following Table 5.
Table 10. The similarity between each alternative to PFPIS and PFNIS
respectively

S md S + ( Gi ) S − ( Gi ) Snd S + ( Gi ) S − ( Gi ) S zd S + ( Gi ) S − ( Gi )

G1 0.9393 0.9776 G1 0.8855 0.9562 G1 0.9067 0.9649


G2 0.9772 0.9768 G2 0.9554 0.9546 G2 0.9643 0.9637
G3 0.9964 0.9298 G3 0.9928 0.8688 G3 0.9943 0.8927
G4 0.9964 0.9298 G4 0.8348 0.9224 G4 0.8638 0.9374

Table 11. Relative closeness coefficients for each alternative Di

Smd ( )
S Di Snd ( )
S Di S zd S Di( )
D1 0.4900 D1 0.4808 D1 0.4844
D2 0.5001 D2 0.5002 D2 0.5001
D3 0.5173 D3 0.5333 D3 0.5269
D4 0.4867 D4 0.4751 D4 0.5269

The results of ranked alternatives are exhibited in Table 12.

Zahid Hussain et al

82
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86
Table 12. The ranking order of alternatives for different entropies
Similarity Ranking Best alternative
S md D3 D2 D1 D4 D3
Snd D3 D2 D1 D4 D3
S zd D3 D2 D1 D4 D3

The ranking results in Table 12 show that our proposed similarity measures Eqs. [X-
XII] rank the alternative in similar order and there is no conflict in ranking the
alternative. Completely, numerical assessment outcomes and application clarify that
the suggested similarity measures Eqs. [X-XII] is much simpler and well suitable in a
multi-criterion decision-making environment. Clearly, the alternative D3 is best
among all the alternatives.

V. Conclusion

In this paper, we use divergence measure to construct the difference between


two Pythagorean fuzzy sets. We studied the axiomatic definition of divergence
measures and their properties. To illustrate the applicability of the proposed measure,
we have derived three similarity measures and then utilized them in pattern
recognition, characterization of the similarity between linguistic variables and multi-
criteria decision-making problems. We have examined the analysis of numerical
results obtained from our proposed similarity measures Eqs. [X-XII] very carefully to
check the performance and behavior in different environments, we gave some
examples in pattern recognition and linguistic hedges like “large”, “more or less’’,
“very large” and “very very large”, which might be suitable and useful to decide
fuzziness of Pythagorean fuzzy sets more clearly and meaningfully. Finally, we have
applied our proposed similarity measures Eqs. [X-XII], respectively in daily life
problem to rank Belt and Road initiative (BRI) projects under the umbrella of expert
opinion involving multi-criterion decision making. On the basis of numerical
simulation results obtained, we observed that our proposed similarity measures Eqs.
[X-XII] are straightforward, sensible, intuitive and well suitable in conduct different
types of problems relating to pattern recognition, linguistic variables and multicriteria
decision making in the Pythagorean fuzzy environment.

Conflict of interest
All the authors are hereby declared that they have no conflict of interest.

Zahid Hussain et al

83
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86
References

I. Abbas, S. E. (2005). On intuitionistic fuzzy compactness. Information


Sciences, 173(1-3), 75-91.
II. Atanassov, K. T. (2000). Two theorems for intuitionistic fuzzy
sets. Fuzzy sets and systems, 110(2), 267-269.
III. Atanassov K, (1999). Intuitionistic Fuzzy Sets, Theory and
Applications, Heidelberg: Physica-Verlag, New York.
IV. Atanassov K,(1994a). New operations defined over the intuitionistic
fuzzy sets. Fuzzy SetsSyst, 61,137–142.
V. Atanassov K, (1989). More on intuitionistic fuzzy sets. Fuzzy Sets Syst,
33,37–46.
VI. Atanassov K, & Gargov G,(1989). Interval valued intuitionistic fuzzy
sets. Fuzzy Sets Syst,31,343–349.
VII. Atanassov K,(1986). Intuitionistic fuzzy sets. Fuzzy Sets Syst. 20:87–
96.
VIII. Bouchon-Meunier, B., Rifqi, M., & Bothorel, S. (1996). Towards
general measures of comparison of objects. Fuzzy sets and
systems, 84(2), 143-153.
IX. Candan, K. S., Li, W. S., & Priya, M. L. (2000). Similarity-based
ranking and query processing in multimedia databases. Data &
Knowledge Engineering, 35(3), 259-298.
X. Garg, H. (2016). A new generalized Pythagorean fuzzy information
aggregation using Einstein operations and its application to decision
making. International Journal of Intelligent Systems, 31(9), 886-920.
XI. Gou, X., Xu, Z., & Ren, P. (2016). The properties of continuous
Pythagorean fuzzy information. International Journal of Intelligent
Systems, 31(5), 401-424.
XII. Hussian, Z., & Yang, M. S. (2019). Distance and similarity measures
of Pythagorean fuzzy sets based on the Hausdorff metric with
application to fuzzy TOPSIS. International Journal of Intelligent
Systems, 34(10), 2633-2654.
XIII. Hwang, C. M., & Yang, M. S. (2016). Belief and plausibility functions
on intuitionistic fuzzy sets. International Journal of Intelligent
Systems, 31(6), 556-568.
XIV. He, Y., Hamza, A. B., & Krim, H. (2003). A generalized divergence
measure for robust image registration. IEEE Transactions on Signal
Processing, 51(5), 1211-1220.
XV. Kacprzyk, J., & Ziolkowski, A. (1986). Database queries with fuzzy
linguistic quantifiers. IEEE transactions on systems, man, and
cybernetics, 16(3), 474-479.
XVI. Lin, J. (1991). Divergence measures based on the Shannon
entropy. IEEE Transactions on Information theory, 37(1), 145-151.

Zahid Hussain et al

84
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86
XVII. Hwang, C. M., Yang, M. S., & Hung, W. L. (2018). New similarity
measures of intuitionistic fuzzy sets based on the Jaccard index with its
application to clustering. International Journal of Intelligent
Systems, 33(8), 1672-1688.
XVIII. Mitchell, H. B. (2003). On the Dengfeng–Chuntian similarity measure
and its application to pattern recognition. Pattern Recognition
Letters, 24(16), 3101-3104.
XIX. Montes S, Couso I, Gil P, &Bertoluzza C,(2002). Divergence between
fuzzy sets. Int. J. of Approx. Reasoning, 31,91–105.
XX. Peng, X., & Garg, H. (2019). Multiparametric similarity measures on
Pythagorean fuzzy sets with applications to pattern
recognition. Applied Intelligence, 49(12), 4058-4096.
XXI. Peng, X., & Yang, Y. (2017). Algorithms for interval-valued fuzzy soft
sets in stochastic multi-criteria decision making based on regret theory
and prospect theory with combined weight. Applied Soft
Computing, 54, 415-430.
XXII. Papageorgiou, E. I., & Iakovidis, D. K. (2012). Intuitionistic fuzzy
cognitive maps. IEEE Transactions on Fuzzy Systems, 21(2), 342-354.
XXIII. Petry FE,(1996). Fuzzy Database: Principles and Applications. Kluwer,
Dordrecht.
XXIV. Pal, N. R., & Pal, S. K. (1992). Some properties of the exponential
entropy. Information sciences, 66(1-2), 119-137.
XXV. Pal, N. R., & Pal, S. K. (1991). Entropy: A new definition and its
applications. IEEE transactions on systems, man, and
cybernetics, 21(5), 1260-1270.
XXVI. R. Shashi Kumar Reddy, M.S. Teja, M. Sai Kumar, Shyamsunder
Merugu, : ‘MATHEMATICAL INPUT-OUTPUT RELATIONSHIPS &
ANALYSIS OF FUZZY PD CONTROLLERS’. J. Mech. Cont. & Math.
Sci., Vol.-15, No.-8, August (2020) pp 330-340. DOI :
10.26782/jmcms.2020.08.00031
XXVII. Ren, P., Xu, Z., & Gou, X. (2016). Pythagorean fuzzy TODIM
approach to multi-criteria decision making. Applied Soft
Computing, 42, 246-259.
XXVIII. Rényi, A. (1961). On measures of entropy and information.
In Proceedings of the Fourth Berkeley Symposium on Mathematical
Statistics and Probability, Volume 1: Contributions to the Theory of
Statistics. The Regents of the University of California.
XXIX. Szmidt, E., & Kacprzyk, J. (2009, July). Analysis of Similarity
Measures for Atanassov's Intuitionistic Fuzzy Sets. In IFSA/EUSFLAT
Conf. (pp. 1416-1421).
XXX. Szmidt, E., & Kacprzyk, J. (2000). Distances between intuitionistic
fuzzy sets. Fuzzy sets and systems, 114(3), 505-518.
XXXI. Shannon CE, &Weaver,(1949). The Mathematical Theory of
Communication. University of Illinois,Urbana, 3–91.
XXXII. Tahani, V. (1977). A conceptual framework for fuzzy query
processing—a step toward very intelligent database
systems. Information Processing & Management, 13(5), 289-303.
Zahid Hussain et al

85
J. Mech. Cont. & Math. Sci., Vol.-16, No.-6, June (2021) pp 64-86
XXXIII. Wu, K. L., & Yang, M. S. (2002). Alternative c-means clustering
algorithms. Pattern recognition, 35(10), 2267-2278.
XXXIV. Xiao, F., & Ding, W. (2019). Divergence measure of Pythagorean
fuzzy sets and its application in medical diagnosis. Applied Soft
Computing, 79, 254-267.
XXXV. Xia, M., Xu, Z., & Liao, H. (2012). Preference relations based on
intuitionistic multiplicative information. IEEE Transactions on Fuzzy
Systems, 21(1), 113-133.
XXXVI. Xu, Z. (2007). Intuitionistic preference relations and their application
in group decision making. Information sciences, 177(11), 2363-2379.
XXXVII. Yang, M. S., & Hussain, Z. (2018). Fuzzy entropy for pythagorean
fuzzy sets with application to multicriterion decision
making. Complexity, 2018.
XXXVIII. Yager, R. R. (2013). Pythagorean membership grades in multicriteria
decision making. IEEE Transactions on Fuzzy Systems, 22(4), 958-
965.
XXXIX. Yager, R. R., & Abbasov, A. M. (2013). Pythagorean membership
grades, complex numbers, and decision making. International Journal
of Intelligent Systems, 28(5), 436-452.
XL. Yager, R. R. (2013, June). Pythagorean fuzzy subsets. In 2013 joint
IFSA world congress and NAFIPS annual meeting (IFSA/NAFIPS) (pp.
57-61). IEEE.
XLI. Yang, M. S., & Wu, K. L. (2004). A similarity-based robust clustering
method. IEEE Transactions on Pattern Analysis and Machine
Intelligence, 26(4), 434-448.
XLII. Zhou, Q., Mo, H., & Deng, Y. (2020). A new divergence measure of
pythagorean fuzzy sets based on belief function and its application in
medical diagnosis. Mathematics, 8(1), 142.
XLIII. Zhang, X. (2016). A novel approach based on similarity measure for
Pythagorean fuzzy multiple criteria group decision
making. International Journal of Intelligent Systems, 31(6), 593-611.
XLIV. Zhang, S. F., & Liu, S. Y. (2011). A GRA-based intuitionistic fuzzy
multi-criteria group decision making method for personnel
selection. Expert Systems with Applications, 38(9), 11401-11405.
XLV. Zadeh, L. A. (1971). Similarity relations and fuzzy
orderings. Information sciences, 3(2), 177-200.
XLVI. Zadeh LA, (1965). Fuzzy sets. Information and Control, 8,338–353.

Zahid Hussain et al

86

You might also like