The Selection of the Dempster’s Rule Based on Evidence Trap Problem
in D-S Theory
Haixin Zhang, Bingyi Kang, Daijun Wei, Ya Li, Juan Liu and Yong Deng
Abstract— Dempster’s rule may not handle the conflicting
belief structures in several situations. The Dempster’s combi-
nation rule and its alternatives have been under the microscope.
This paper focus on the conflicting belief structure problem and
comes up with an evidence trap problem. A method of deter-
mining whether to select the Dempster’s rule of combination
or not when there is an evidence trap is proposed. The method
deploys the conflict coefficient and the distance between belief
structures as a two-dimensional measure with two thresholds
settled. Numeric examples show the efficiency of the proposed
method.
Index Terms— Evidence trap, D-S theory, Conflict, Alterna-
tive combination rule
I. INTRODUCTION
The Dempster-Shafer theory of evidence (D-S theory)[1],
[2] provides a very useful framework for representing and
manipulating information, represented as different belief
structures from different sources, especially in environments
with some uncertainties and ignorance[3], [4], [5]. A central
component called Dempsters rule of the reasoning process
of this methodology is the aggregation of multiple belief
structures. However criticisms on the counterintuitive results
of applying Dempsters combination rule to post-called con-
flicting belief structures soon emerged since this methodol-
ogy first appeared, where an almost impossible choice (with
a very lower degree of belief) by both sources came up
as the most possible outcome (with a very high degree of
belief)[6], [7]. In other words, applying the Dempster’s rule
to the belief structures will produce an unreasonable result
in some situations. Confused by the unreasonable results,
many researchers have studied the conflicting belief structure
problem, and some proposed to modify the combination
rule and several alternative combination rules in D-S theory
have been proposed[8], [9], [10],while others proposed to
transform the belief structures[11], [12]. The issue of the
management of conflicts between pieces of knowledge is a
fundamental problem in the development of all automated
reasoning systems and due to the increasing applications
of D-S theory in intelligent fusion processes, Dempsters
The work is partially supported National Natural Science Foundation of
China, Grant No.60874105, 61174022, Program for New Century Excellent
Talents in University, Grant No.NCET-08-0345, Chongqing Natural Science
Foundation, Grant No. CSCT, 2010BA2003, the Fundamental Research
Funds for the Central Universities Grant. No XDJK2010C030, Grant.
No XDJK2011D002, Doctor Funding of Southwest University Grant No.
SWU110021. The first author also greatly appreciates the support by the
School of Computer and Information Sciences of Southwest University
Scientific and Technological Innovation Fund for Students.
{Haixin Zhang, Bingyi Kang, Daijun Wei, Ya Li, Juan Liu and Yong
Deng} are with the School of Computer and Information Sciences, South-
west University, Chongqing, 400715, China
ydeng@swu.edu.cn
combination rule and its alternatives have been under the
microscope. This paper focus on how to distinguish the
conflicting belief structures and suggested a way to determine
whether select the Dempsters rule of combination or not
when facing an evidence trap.
The remaining paper is is organized as follows. Section 2
reviews the basic D-S theory and defines what an evidence
trap is. The conflict is analyzed and discussed, and a method
to judge whether to apply the Dempsters combination rule is
proposed in section 3, which deploys quantitative measures
of both the conflict coefficient and the distance between
BPAs. Brief conclusions is drawn in section 4.
II. EVIDENCE TRAP DESCRIPTION
A. Basic D-S theory
In this subsection we briefly review the D-S theory of
evidence. More comprehensive and detailed information can
be found in the literature[1], [2].
Definition 2.1: Let Ω be the frame of discernment, with
N exclusive and exhaustive solutions {θ
1
,θ
2
, ...θ
N
}, denoted
Ω={θ
1
,θ
2
, ...θ
N
}. A basic probability assignment (BPA)
is a mapping m :2
Ω
→ [0, 1] that satisfies
m(A
i
) > 0
A⊆Ω
m(A
i
)=1,
where m (∅)=0is required.
A belief structure m consists of a collection of non-null
subsets, A
1
,A
2
, ...A
n
of Ω called focal elements and a set
of weights m(A
i
), often be represented by BPA, and the
mapping is also called the mass function. m(A) defines the
amount of belief to the subset A exactly, not including any
subsets in A. The total belief in a subset A is the sum of all
the mass assigned to all subsets of A. This function is known
as a belief function and is defined as Bel :2
Ω
→ [0, 1].
Bel (A)=
B⊆A
m (B) (1)
When m(A) > 0, A is referred to as a focal element of
the belief function. A plausibility function, denoted Pl,is
defined as follows, where Pl :2
Ω
→ [0, 1].
Pl(A)= 1−Bel
A
=
A∩B=∅
m (B) (2)
where
A is the complementary set of A.
We can see Pl(A) as the expected possibility over the
focal elements and Bel(A) is the expected certainty. In[1],
[15] Dempster shows that these two measures play an impor-
tant role in the determination of the probability of a subset
3122
978-1-4577-2074-1/12/$26.00
c
2012 IEEE