AUTHOR COPY
2280 X. Zhang and Z. Xu / Novel distance and similarity measures on hesitant fuzzy sets
condition that 0 ≤ µ
ξ
(x
i
),µ
ζ
(x
i
) ≤ 1, for x
i
∈ X
(i = 1, 2,...,n); w
i
is the weight of the element
x
i
∈ X satisfied the condition that w
i
∈ [0, 1] and
n
i=1
w
i
= 1.
With the growing complexity and uncertainty of
the real-life problems, many different extensions of
fuzzy sets, such as intuitionistic fuzzy sets (IFSs) [2],
interval-valued IFSs [1], linguistic variables [34], type-
2 fuzzy sets [15], etc., were proposed to model such
uncertainty inherent of these problems. Correspond-
ingly, lots of the extensions of these aforementioned
distance measures have been developed for measur-
ing the distances between these extensions of fuzzy
sets, such as IFSs [8–14, 18, 22, 29], interval-valued
IFSs [23, 26], type-2 fuzzy sets [24, 32] and linguistic
variables [27], etc. For instance, by taking three param-
eters of IFS into account, Szmidt and Kacprzyk [18]
proposed some intuitionistic fuzzy distance measures
between IFSs. Li and Cheng [11] generalized the Ham-
ming and Euclidean distances by adding a parameter
and developed an intuitionistic fuzzy similarity mea-
sure for IFSs. Afterwards, some authors [13, 14, 22]
improved Li and Cheng’ method [11] to develop some
new similarity measures for IFSs. Based on the Haus-
dorff metric, Grzegorzewski [8], Hung and Yang [9]
proposed a series of similarity measures for IFSs and
interval-valued fuzzy sets. Xu and Chen [29] made
a comprehensive overview of distance and similarity
measures for IFSs whereby they also developed several
intuitionistic fuzzy continuous distance and similar-
ity measures. Yang and Lin [32] defined the similarity
and inclusion measures between type-2 fuzzy sets. Wu
and Mendel [24] proposed a new similarity measure
for interval type-2 fuzzy sets. In the linguistic fuzzy
context, Xu [27] introduced the concepts of deviation
degrees and similarity degrees between two linguistic
variables.
Recently, another interesting extension of fuzzy set,
i.e., hesitant fuzzy set (HFS), has been developed by
Torra [20] to manage the practical situations in which
one hesitates among several possible values to assess
an indicator, alternative, variable, etc. Basic elements
in HFSs are hesitant fuzzy elements (HFEs) [25] which
include some possible values for the membership of an
element to a set. For example, to discuss the member-
ship of x to a set A, a decision organization including
several experts may hesitate among some possible val-
ues as 0.3, 0.5 and 0.6, then the hesitance experienced
by the decision organization can be modeled by an HFE
{0.3,0.5,0.6}. It is noted that the HFE {0.3,0.5,0.6}
can describe the above situation more objectively than
the crisp number 0.3 (or 0.5, or 0.6), or the inter-
val value [0.3, 0.6], or the intuitionistic fuzzy value
(0.3, 0.4), because the membership is not the convex
of 0.3 and 0.4, or the interval between 0.3 and 0.6, but
just three possible values [30]. Thus the use of hesi-
tant fuzzy assessments makes the people’s judgments
more reliable and informative in practical applications
[17, 36, 37]. To apply the HFSs into various sci-
entific fields (such as pattern recognition, clustering
analysis, etc.) more deeply, Xu and Xia [30] pro-
posed a variety of hesitant fuzzy distance measures
and ordered distance measures for HFSs, and discussed
their properties. Based on Xu and Xia’s hesitant fuzzy
distance and similarity measures, Farhadinia [7] investi-
gated the relationship among the entropy, the similarity
measure and the distance measure for HFSs and mean-
while extended these measures into interval-valued
HFSs.
However, it is worth mentioning that the Xu and Xia’s
hesitant fuzzy distance measures [30] do not fit well in
some cases. For example, let M, N, and Q be HFSs
in X, where M ={<x
i
, {0.3, 0.5} > |x
i
∈ X}, N =
{<x
i
, {0.2, 0.3} > |x
i
∈ X}, Q ={<x
i
, {0.2, 0.7} >
|x
i
∈ X}. Using Xu and Xia’s hesitant fuzzy distance
measures we can obtain that the distance between M
and N is equal to the distance between M and Q.
It is noted that we cannot carry out our comparison
in such a case using the Xu and Xia’s hesitant fuzzy
distance measures. This will get the decision-makers
into trouble in practical applications. To overcome this
drawback, this study proposes a novel concept of hes-
itancy index of HFS to measure the hesitant degree
among the possible values in each HFE of each HFS.
By taking the hesitancy indices into account, a series
of novel hesitant distance and similarity measures for
measuring hesitant fuzzy information are developed.
It can be easily seen that the hesitant degrees of M,
N, and Q are different because the decision-maker
hesitates between 0.3 and 0.5 when providing the mem-
bership of x to a set M, while hesitates between 0.2
and 0.3 to N as well as between 0.2 and 0.7 to Q.
Comparing with the Xu and Xia’s distance and sim-
ilarity measures [30], we find that the new distance
and similarity measures proposed in this study are more
reasonable.
Owing to the fact that distance and similarity mea-
sures are fundamentally important in the clustering
analysis fields, we further apply the novel hesi-
tant fuzzy similarity measures in clustering analysis
under hesitant fuzzy environments and develop a new