A set X # U with mðXÞ – 0 is called a focal element of m. Let M be a family of all focal elements of m. A pair of ðM; mÞ is called
a belief structure on U.
Definition 2.2. The belief function is a mapping Bel:2
U
!½0; 1 that satisfies
BelðXÞ¼
X
X
0
# X
mðX
0
Þ;
8
X 2 2
U
:
Definition 2.3. The plausibility function is a mapping Pl:2
U
!½0; 1 that satisfies
PlðXÞ¼
X
X\X
0
–/
mðX
0
Þ;
8
X 2 2
U
:
According to the belief structure, a pair of belief and plausibility functions can be derived. Based on the same basic proba-
bility assignment, the belief and plausibility functions are dual, i.e., BelðXÞ¼1 Plð XÞ, where X is the complement of X.
½BelðXÞ; PlðXÞ is the confidence interval which describes the uncertainty about X, and PlðXÞBelðXÞ represents the level of
ignorance about X. Furthermore, a belief function satisfies the following axioms:
(1) Belð;Þ ¼ 0,
(2) BelðUÞ¼1,
(3) Belð[
m
i¼1
X
i
Þ P
P
;–J # f1;2;...;mg
ð1Þ
j Jjþ1
Belð\
i2J
X
i
Þ for any X
1
; X
2
; ...; X
m
# U.
When uncertain information comes from different sources, it is important to desire a consensus by combining such infor-
mation. Dempster–Shafer theory uses Dempster’s combination rule for combining belief functions defined by independent
bodies of evidence.
Definition 2.4. Let m
1
and m
2
be two bpas on U which are derived from two distinct sources. Then, the combination (or the
joint) m
12
is calculated from the aggregation of m
1
and m
2
in the following manner:
m
12
ðXÞ¼
P
B\C¼X
m
1
ðBÞm
2
ðCÞ
1 k
;
8
X 2 2
U
; X – ; ;
where k ¼
P
B\C¼;
m
1
ðBÞm
2
ðCÞ and represents the basic probability mass associated with conflict. This is determined by the
summing the products of the bpa’s of all sets where the intersection is null. In general, we denote m
12
ðXÞ by m
1
m
2
ðXÞ
which is the so-called orthogonal sum operator of combination.
2.3. Qualitative fusion functions based on multigranulation rough sets
According to two different approximating strategies, i.e., seeking common reserving difference and seeking common
rejecting difference, Qian and Liang [23] proposed two kinds of multigranulation rough sets which provide two kinds of dif-
ferent lower or upper fusion functions. They are optimistic lower and upper multigranulation fusion functions and pes-
simistic versions, which are disjunctive and conjunctive combination rules, respectively.
Definition 2.5. Let MS ¼fIS
i
jIS
i
¼ðU; AT
i
; fðV
a
Þ
a2AT
i
g; f
i
Þg be a multi-source information system. K
1
; K
2
; ...; K
q
are q granular
structures induced by AT
1
; AT
2
; ...; AT
q
and X # U. The optimistic lower and upper approximations of X in a multigranulation
rough set can be formally represented as two qualitative fusion functions, respectively,
X
q
i¼1
K
i
o
ðXÞ¼f
o
l
ðK
1
; K
2
; ...; K
q
Þ¼fx 2 Uj½x
K
1
# X __½x
K
q
# Xg;
X
q
i¼1
K
i
o
ðXÞ¼f
o
u
ðK
1
; K
2
; ...; K
q
Þ¼
X
q
i¼1
K
i
o
ð XÞ;
where f
o
l
is called an optimistic lower qualitative fusion function, and f
o
u
is called an optimistic upper qualitative fusion func-
tion. They are a kind of disjunctive combination rules. These two functions are used to compute the lower and upper approx-
imations of a multigranulation rough set through fusing q granular structures. In practical applications of multigranulation
rough sets, the fusion function has many different forms according to various semantics and requirements. It is noted that
the essence of a so-called optimistic lower (or upper) approximation defined in Definition 2.5 is the same as weak lower (or
strong upper) approximation proposed in [7,8].
G. Lin et al. / Information Sciences 314 (2015) 184–199
187