Received January 25, 2019, accepted February 18, 2019, date of publication February 27, 2019, date of current

version August 16, 2019.

Digital Object Identifier 10.1109/ACCESS.2019.2901932

The Negation of Basic Probability Assignment

XIAOZHUAN GAO AND YONG DENG
Institute of Fundamental and Frontier Science, University of Electronic Science and Technology of China, Chengdu 610054, China
Corresponding author: Yong Deng (dengentropy@uestc.edu.cn)
This work was supported by the National Natural Science Foundation of China under Grant 61573290 and Grant 61503237.

ABSTRACT In many cases, we obtain information using various methods in order to make better decisions.
The everything in nature and society has its negative, the negation of negation has significant meaning.
Considering the problem from two aspects, we can get more accurate information. However, in most cases,
the information of negation is ignored. Hence, the negation provides a new view to obtain information.
However, existing negation method mainly apply to probability distribution. How to get the negation of basic
probability assignment (BPA) in Dempster-Shafer (D-S) theory is still an open issue. The paper proposed
the new negation method of BPA. Besides, some numerical examples are given to this approach for better
understanding. Moreover, in order to demonstrate the efficiency of the proposed method, the paper compared
the changes of uncertainty between original and negation by using some uncertain measurement methods.
Finally, practical application is used to be discussed the application of proposed method.

INDEX TERMS Dempster-Shafer evidence theory, negation, basic probability assignment, belief function,
uncertainty.

I. INTRODUCTION
Normally, information is expressed via a degree of impreci-
sion. A large number of approaches have been developed to
solve the problem of expressing the knowledge contained in
sources of information, such as, fuzzy sets [1]–[4], Dempster-
Shafer (short for D-S) theory [5], [6], and rough sets [7],
Z number [8], [9] R number [10], [11] and so on [12]–[14].
Besides, D-S theory as an important and widely used uncer-
tain reasoning method [15]–[17], appeals to many scientists
to search [18]–[21]. D-S theory assigns probabilities to the
power set of events, so as to better grasp the unknown and
uncertainty of the problem [22], [23]. Based on the advan-
tages of D-S theory, the paper mainly discusses the related
contents of D-S theory.
In most cases, if it is known that p(A) = 0, it is must A is
not our choice, hence, we can get some information to make
decisions from known conditions [24], [25]. For example,
to express our knowledge in a numerical way, if a < b,
it can be got −a > −b. From the above example, we can
get the negative information by known information. Besides,
the negation method provides a way to represent the informa-
tion in an opposite side, thus we can get more information to
present knowledge. Moreover, everything in both nature and
The associate editor coordinating the review of this manuscript and
approving it for publication was Xin Luo.
society has its negation side, the negation of everything has
some significant meaning. Hence, it is reasonable to provide
the method of negation to observe and study the knowledge
representation. Besides, Zhade in his BISC blog raised the
negation of a probability distribution. Negation operation
provides a possible way of knowledge representation [26].
For most cases, it is much easier to describe the negation of
the things than directly describe it in some circumstances. For
example, if it is difficult to prove a mathematical formula
rigorously, however a counterexample can easily prove it
wrong. Smets [27] mentioned that the negation of a belief
function could be defined by the implicity function and com-
monality function. The properties of negation of a probability
distribution recently defined by Yager [26] are studied, and
the negation model is proved to have the maximum entropy
allocation.
The negation can not only provide a new way to obtain
information from another view, but also is promising in some
applications, such as non-commutative fuzzy structures [28],
MV-algebras [29], fuzzy logic [30] and involutive [31]. For
example, Bell explicitly incorporated regret, expected utility
theory not only becomes a better descriptive predictor but
also may become a more convincing guide for prescrib-
ing behavior to decision makers [32]. In addition, negation
method is promising in muti-criteria decision (MCDM) mak-
ing, for example, using an ideal solution and the negative

2169-3536
2019 IEEE. Translations and content mining are permitted for academic research only.
107006 Personal use is also permitted, but republication/redistribution requires IEEE permission. VOLUME 7, 2019
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
X. Gao, Y. Deng: Negation of BPA

ideal solution can provide some method to make decision. which satifies
The best alternative is as close as the ideal solution and is as
far as the negative ideal solution. Based on above discussion, m (φ) = 0 (4)
X
it can be seen that providing the negation of BPA is of great m (A) = 1 (5)
significance to study uncertainty. Besides, there are many A⊆2
researches to study the methods of negation [33]–[35]. This In D-S theory, a mass function is also called a basic probabil-
paper is to find a more general method to obtain the negation ity assignment (BPA) or a piece of evidence or belief struc-
of the BPA. Besides, the paper also discussed what are there ture [65]. The m(A) measures the belief exactly assigned to A
the changes after taking negation? What is the cause of these and represents how strongly the piece of evidence supports A.
changes? In physics, some operations can cause change of If m (A) > 0, A is called a focal element, and the union
entropy, which is a measurement of the degree of chaos [9], of all focal elements is called the core of a mass function
[36]. With the increasing application of entropy, informa- [66]. Recently, some operations on BPA, such as correlation
tion entropy has become an indispensable part of modern analysis [67], have been paid attention.
scientific development [37], [38]. So the paper use entropy Definition 3 (Belief Function): The belief function (Bel)
of random sets [39] to measure the changes of entropy after [6] is a mapping from set 22 to [0,1] and satisfied:
taking negation. X
The rest of this article is structured as follows: In Bel (A) = m (B) (6)
Section 2, some preliminaries of D-S evidence theory B⊆A
and some measurements of uncertainty are introduced.
Definition 4 (Plausibility Function): The plausibility
Section 3 details Yager’s negation and proposed negation
function (Pl): 22 → [0, 1], and satisfied [6]:
method, besides, discusses two theorem of proposed negation X
method. In section 4, some numerical examples are used to Pl (A) = m (B) = 1 − Bel(Ā) (7)
illustrate proposed method and discuss some changes after B∩A6=φ
negation. Section 5 discusses the application of proposed
method. Finally, some conclusions are given in Section 6. where Ā = 2 − A. As can be seen from the above, ∀A ⊆ 2,
Bel(A) < Pl(A), Bel(A), Pl(A) are respectively the lower and
II. PRELIMINARIES upper limits of A, namely [Bel(A), Pl(A)], which indicates
A. DEMPSTER-SHAFER THEORY uncertain interval for A and is essential for D-S evidence
The real world is very complicated with uncertainty theory [68]. According to Shafer’s explanation [5], the dif-
[40]–[42]. D-S theory offers a useful fusion tool for uncer- ference between the belief and the plausibility of a propo-
tain information [18], [43]–[45]. D-S theory needs weaker sition A expresses the ignorance of the assessment for the
conditions than the Bayesian theory of probability, so it is proposition A.
often regarded as an extension of the Bayesian theory, which Definition 5 (Conflict): The conflict of two evidence was
has been used to many applications, such as making-decision defined as follows [69]:
[46]–[49], target recognition [50], pattern classification [51],
X
C= m1 (Ai ) × m2 (Bj ) (8)
[52], uncertainty model [53]–[55], conflict management [56], Ai ∩Bj =φ
[57] etc [58]–[60]. Besides, some new or generalized model
based this theory have been proposed, for example, gener- Assume frame of discernment 2, given the two evidences
alized evidence theory [61], [62] and D numbers [63], [64]. E1 and E2, the mass functions are m1 and m2 respectively.
Some preliminaries in D-S theory are introduced as follows. Conflict C is an essential tool in information fusion, many
For additional details about D-S theory, refer to [5], [6]. researches studied it [70]–[72], besides, C ⊂ [0, 1].
Definition 1 (Frame of Discernment): Let 2 be the set of Definition 6 (Dempster’s Rule of Combination): Assume
mutually exclusive and collectively exhaustive events Ai [5], there are two BPAs indicated by m1 and m2 , the Demp-
[6], namely ster’s rule of combination is used to combine them as
follows [5], [6]:
2 = {A1 , A2 , · · · , An } (1)
m (B) × m2 (C) , A 6 = φ
 1 P 
The power set of 2 composed of 2N elements of is indicated m (A) = 1−C B∩C=A 1 (9)
0, A=φ
by 22 , namely [5], [6]:
It is noted that the Dempster’s rule of combination
22 = {φ, {A1 } , {A2 } , · · · , {A1 , A2 } , · · · , 2} (2) is only applicable to such two BPAs which satisfy the
condition C < 1.
Definition 2 ( Mass Function): For a frame of discernment
2 = {A1 , A2 , · · · , An }, the mass function m is defined as a
B. UNCERTAINTY MEASUREMENT
mapping of m from 0 to 1 [5], [6], namely:
Since the uncertainty is extremely important in theory of
m : 22 → [0, 1] (3) information, many people have attempted to understand it

VOLUME 7, 2019 107007

 X. Gao, Y. Deng: Negation of BPA

better through a study of axiom systems which characterize B. PROPOSED NEGATION

information measures [73]–[76]. Interestingly, BPA is similar to probability, but it can express
Definition 7 (Entropy of Random Sets): The entropy of imprecision. Thus, BPA also has its original side and the
random sets was defined as follows [39]: negative side. Now, the procedures of taking negation are
X listed as follows.
Hn = − m(a) × log2 m(a) (10)
Assume frame of discernment 2 has N elements, 22 can be
Definition 8 (Confusion Measure): Confusion measure expressed as:
(Conf) [77] is defined by:
22 = A1 , A2 , · · · , A2N

X (17)
Conf (m) = − m(A)log2 (Bel(A)) (11)
Especially, A1 = φ,A2N = 2. Let m be a BPA. Assume that
Definition 9 (Dissonance Measure): Dissonance measure
2 N
(Diss) [78], is indicated by X
X mAi = 1 (18)
Diss(m) = − m(A)log2 (Pl(A)) (12) i=1
Definition 10 (Measures for Non-Specificity): Measures Similarly, m̄Ai represent the negation of mAi , and the procedure
for Non-Specificity (NS) [78], is indicated by of taking negation is listed as follows.
Step1 : For each focal element Ai in the frame of dis-
X
NS(m) = m(A)log|A| (13)
A⊂2
cernment, we use 1 − mAi to represent the complementary
probability of mAi .
where |A| is the cardinality of A. In D-S theory, NS indicates
the uncertain degree of a BPA which contains two or more m̄Ai = 1 − mAi (19)
unspecified alternatives [33]. NS is a kind of generation
of Hartley measure [79] from the classical set theory [80]. Step2 : Calculate the sum λ of the negative belief of all the
Besides, it only focuses on those focal elements with the focal elements. Namely,
larger cardinality.
X
λ= m̄Ai (20)
III. NEGATION METHOD Assume 2 = {a, b}, the BPA is represented as m(a) = 0.3,
By analysing, it is known that negation has been used in m(b) = 0.2, m(a, b) = 0.5. The negation of the mass function
many applications and give us additional information, for m̄ is
example, expect provide information, while regret which is
m̄(a) = 0.7, m̄(b) = 0.8, m̄(a, b) = 0.5
the negation of expect can provide additional information.
Besides, if a probability distribution contains event with low So, it can be calculated the sum of negation of the mass
probabilities or zero probabilities, then its negation is bound function: λ = 2 = 22 − 2. The specific mathematical proof
to represent events with high probabilities. Yager [26] observe is as follows.
that the probabilities in the negation of a probability distri- If the frame of discernment is unknown, for a mass function
bution are bounded in an interval (unless it is a Bernoulli m (a) = a, m (b) = b, m (a, c) = c, then:
distribution where a probability distribution and its negation
will be identical). In the following we introduce a negation of m̄ (a) = 1 − a; m̄ (b) = 1 − b; m̄ (a, c) = 1 − c;
probability distribution. m̄ (a)+ m̄ (b) + m̄ (a, c) = 2

A. YAGER’S NEGATION It can be known that the power set of frame of discernment
Consider a probability distribution: composed of 2n elements, so there is 2n − 4 except for
mass function is 0. Besides, the corresponding negation is
n
( )
X 1 respectively, hence, the sum of negation is 2n − 4. From the
p = p1 , p2 , · · · , pn ; 0 6 pi 6 1; pi = 1 (14)
above, it can be seen that the sum of negation of mass function
i=1
does not equal 1, it should be λ = 2n − 4 + 2 = 2n − 2. So,
defined on the the set X = (x1 , x2 , · · · , xn ). The negation of it is necessary to normalize.
the probability distribution P is defined by the following: Step3 : The third step is used to normalize the belief of all
n
( )
X the negative focal elements.
p̄ = p̄1 , p̄2 , · · · , p̄n ; 0 6 p̄i 6 1; p̄i = 1 (15)
1 − mAi
i=1 m̄Ai = (21)
where λ
Pn From the above, it has known the λ does not equal 1. By calcu-
1 − pi 1 − pi j=1,i6=j pj
p̄i = = Pn = (16) lation,the λ equals 2n − 2. Hence, the negation of BPA should
n−1 j=1 (1 − p̄i ) n−1 be:
where n is the amount of focal element and pi is the probabil- 1 − mAi
ity of original mass function. m̄Ai = n (22)
2 −2
107008 VOLUME 7, 2019
X. Gao, Y. Deng: Negation of BPA
Especially, this general negation method can definitely
degenerate to the basic negation method which is proposed by
Yager if the belief is only assigned to sole element. Indicate
as:
m̄Ai = 1 − mAi (23)
Specific examples are as follows:
Example There is a frame of discernment 2 = {a}, for a
mass function m (a) = 1, the negation is as follows:
m {a} = 0
which is the same with Yager’s negtaion.
In fact, the negation method acts in a manner to assign the
belief evenly to each focal element. There is an example to
explain the proposed negation, as follows.
Example Assume a frame of discernment 2 = {a, b}, for
a mass function m (a) = 0.3, m (b) = 0.2, m (a, b) = 0.5,
the associated negation can be calculated as follows.
m {a} = 0.35, m {b} = 0.4, m {a, b} = 0.35,
Next, let us introduce two general theorems of the proposed
negation method. The Theorem 1 points out the general for-
mula of the negation method, and the Theorem 2 discovers
the general convergence process of the negation method.
Theorem 11: Treating the iterative negations as the series
of sequence, let the xi denote the value of the focal element
after taking the ith negation. The general formula of the
proposed negation method is:
Pt=i−3 n
1 (2 − 1)t − (1 − x1 )
xi = n − t=1 (24)
2 −1 (2n − 1)i−1
Theorem 12: For a BPA m, it converges to its limitation
when the belief is equally assigned to each focal element.
Namely:
1 m̄ (a, b) = 0.5
m(A2 ) = m(A2 ) = · · · = m(A2N ) = (25)
2N − 1
Proof : To simplify our notation, let xi replace m(An ), from
the Theorem 1, the general formula of xi has been obtained.
Pt=i−3 n
1 (2 − 1)t − (1 − x1 )
xi = n − t=1
2 −1 (2n − 1)i−1
Next, taking the limitation on both sides of the equation,
the convergence can be got. Namely:
Pt=i−3 n
1 (2 − 1)t − (1 − x1 )
lim xi = lim n − t=1
i→∞ i→∞ 2 − 1 (2n − 1)i−1
Thus, with the condition that (2n − 1) > 1, it can be obtained:
1 elements.
lim xi =
i→∞ 2n − 1
Next, the special case is discussed in the following part.
Special Case : Assumed a BPA that contains only one focal
element (e.g. the m(a) = 0), the negation of BPA can be
defined by:
m(ā) = 1
VOLUME 7, 2019
From the above, it can be found that the element can not
be a due to m(a) = 0. But the negation give another view that
the element can belong to ā. Besides, in this phenomenon,
we can assume φ represents elements expect a, namely open
world.
Due to the lack of complete knowledge, it can not know
whether the world has only a, called open world. This equa-
tion suggests that the empty set φ can also represent some
focal elements. Incompleteness is the important property in
D-S theory. But, the knowledge is complete, the m(a) = 1
represents the inevitability of the event. Therefore, in the open
world, we do not know which specific focal element can be
assigned to, but at least it is not m(a), so unknown elements
are be assigned to the φ. From this view, negation has some
connections with the open world. The special case prove the
effectiveness of negation. Besides, it also demonstrates that
the m(a) = 0 also has some interesting information we can
not know. In this paper, open world is not considered.
The process of taking the negation of a BPA generates a
new and unique BPA that can be seen as being consistent with
the idea of the negation of the initial BPA. In this context,
the result of the special cases is consistent with our intuitions.
IV. EXPERIMENTAL STUDY
In this section, several examples are given to show the how to
obtain the negation of a mass function based on the proposed
negation method.
A. EXAMPLE
Example 1: Assume a Frame of Discernment 2 = {a, b}, for
a mass function m (a) = 0.5, m (b) = 0.5 the associated
negation can be calculated as follows.
m̄ (a) = 0.25
m̄ (b) = 0.25
From the above, the BPA has some changes after the
negation, that is to say, the BPA has been redistributed.
Example 2: Assume a frame of discernment 2 = {a, b}, for
a mass function m (a) = m (b) = m (a, b) = 31 , then:
1
m̄ (a) = m̄ (b) = m̄ (a, b) =
3
It can be seen that the uniform distribution does not change
after taking negations, which is consistent with the Theorem
2. During the some views, the system with uniform distribu-
tion is the most stable and has no changes with any operation.
Hence, taking negation does not change the distribution of
Example 3: There is a frame of discernment 2 = {a, b, c},
for a mass function m (a) = 0.2, m (b) = 0.1, m (a, c) = 0.3,
m (a, b, c) = 0.4, we can get:
m̄ (a) = 0.133, m̄ (b) = 0.15, m̄ (c) = 0.167,
m̄ (a.b) = 0.167
m̄ (a, c) = 0.116, m̄ (b, c) = 0.167, m̄ (a, b, c) = 0.1
107009

TABLE 1. The trend of negation process of multiple elements.
From the above, the BPA of m(c) can not be known, namely
m(c) = 0, but it is very obviously that c also includes some
information and is basic element of frame of discernment,
because of the existing of m(a, c). Similarly, the (b, c) also
has some information due to the existing of m(a, b, c). It can
be seen that initial BPA can not represent the full information.
The information of m(c) and m(b, c) can not know from the
initial BPA, but, it can be obtained some information from
m̄ (c) and m̄ (b, c). It can be proved that negation has the
essential effect with obtaining more unknown the informa-
tion. Besides, negation also provides a new view to consider
world. Although it can be obtained some information after
negation, what happens if taking negation to the m̄ ? Make
the 2nd negation and obtain the results, as follows:
¯ 6 = m(i)
m̄(i) (26)
¯ m(a)
m̄(a) = n 6 = m(a) (27)
(2 − 2)2
From the above, it find that the negation is irreversible. It is
essential to explore the reason of irreversible after negation.
In the next section, some possible reasons can be presented
for this interesting phenomenon.
B. FURTHER DISCUSSION
Uncertainty is ubiquitous in fuzzy systems, and therefore,
how to handle the uncertainty of information is of great
importance [81], [82]. Next, from different views discuss the
uncertainty of negation.
Example 4: Assume N = 3, 2 = {a, b, c}, the initial BPA
m is as follows: m (a) = 0.2, m (c) = 0.15, m (a, b) = 0.15,
m (a, b, c) = 0.5.The result of doing several negation process
to BPA is shown in Fig. 1.
The Fig. 1 is used to illustrate the change of the BPA
after each iteration of negation process. From the simulation
result, it can easily find that the BPA is converging gradually
with the increasing of iterations. At last, the BPA is uniform
distribution after multiple negation, which is consistent with
Therom 2. In other words, ambiguity uncertainty is maxi-
mum if each focal element has the same probability in this
phenomenon [26].
Besides, Table 1 presents the interval of BPA from Bel(A)
to Pl(A). From Table 1 can find that the interval changes after
each iteration of negation process, meaning that the uncer-
107010
X. Gao, Y. Deng: Negation of BPA
FIGURE 1. The BPA trend of negation process of multiple elements.
tainty of BPA has changes after negation. Besides, the interval
of BPA is gradually approached convergence after multi-
ple negations. Amusingly, the interval of m(a, b, c) has no
changes with the negation, due to m(a, b, c) including the
all basic elements that the sum of them equals 1. But, what
changes have taken with the changes of BPA after taking the
negation? Next, considering the Conf , Diss and NS of BPA,
which these measures taken together provide an indication of
the quality of the evidence supplied by a belief structure [77].
1) VIEW FORM CONFUSION DISSONANCE AND NS
From the Fig. 2, it can be found that Conf , Diss, NS measures
are in a state of fluctuation in several previous negations,
which can be seen uncertainty has some change after nega-
tion. Ultimately, Conf Diss and NS are approaching stability
after multiple negations, that is to say, the system is in a
relatively stable state, indicating the BPA is uniform distribu-
tion. Besides, NS focuses on the focal element with the larger
cardinality, when the cardinality is same, the NS changes with
BPA. Moreover, by observing the changes of NS between
original and first negation, it can be found that the BPA has
taken place more great changes in the first negation than
other negations. In addition, the difference between Conf and
Diss reflects the changes of belief interval of BPA. From
Fig. 2, the Conf has more uncertainty than Diss, showing
the Pl has more information than Bel. Moreover, the Conf
and Diss have the same trend of changes after negation. More
VOLUME 7, 2019
X. Gao, Y. Deng: Negation of BPA

TABLE 2. The Nguyen entropy after negation.

FIGURE 2. The degree of uncertainty after negation.

importantly, according the equation(7), the Bel and Pl have

some connections with negation, so the Pl and Bel have the
consistent changes. Similarly, the Conf and Diss have more proof.
changes taking 1st negation in this case. Hence, it is necessary X
Hn (m) = − m(a) × log2 m(a) (28)
to consider what is the reason of this phenomenon? From the X
above illustration, the uncertainty, which takes the negation Hn (m̄) = − m̄(a) × log2 m̄(a)
twice, does not equal to the initial uncertainty, showing that X 1 − m(a) 1 − m(a)
taking negation is irreversible. Besides, the changes of uncer- =− × log2 N (29)
tainty means the changes of information, which maybe con- 2N − 2 2 −2
Hn (m) − H (m̄)
nect with entropy, after all, entropy is irreversible in physics. X
Similarly, negation is also irreversible. =− m(a) × log2 m(a)
X 1 − m(a) 1 − m(a)
− (− N
× log2 N ) (30)
2) VIEW FORM ENTROPY 2 −2 2 −2
X 1 − m(a)
Recalled the entropy, is derived from physics [83], the entropy
X
D0 = − m(a) × log2 m(a) − (−
is also used to denote the degree of discord. The more 2N − 2
chaotic a system is, the more uniform the state distribution 1 − m(a) X
× log2 N ) + λ( m(a) − 1) (31)
is. The energy consumed is hard to be used again, that is say, 2 −2
the process is reversible. With the increasing application of ∂D0 1 1 − m(a)
=− N log2 N
entropy, information entropy has become an indispensable ∂m(a) 2 −2 2 −2
part of modern scientific development [84]–[86]. From the m(a)
− N log2 e − log2 m(a) − log2 e + λ (32)
above, the negation is irreversible, meaning the information 2 −2
consumes is hard to used again. In this paper, using entropy Hn (m̄) > Hn (m) (33)
of random sets to observe the changes of information entropy.
By calculating, it can be easily found that entropy has always From the above, it can be seen the Hn (m̄) > Hn (m), which
keep increase until maximum entropy. From the views can be is consistent with the above experimental results. Hence, our
seen that the proposed negation has maximum entropy after proposed negation cause the increase of entropy of random
multiple negations. More importantly, after multiple nega- sets until the maximum entropy.
tions, the system is approaching stability, at the same time,
entropy has maximum value. According to this phenomenon, 3) VIEW FORM CONFLICT
it can be considered that no operation will reduce the entropy Conflict plays an important property in D-S theory [87].
in an isolated system according to second law of thermody- Furthermore, the negation always shares the same frame of
namics. Hence, using entropy of random sets to measure the discernment with the initial BPA since the negation function
entropy of BPA is reasonable. Besides, Conf , Diss and NS has only changes the belief distribution instead of the frame
more fluctuation after 1st negation in Example 4, showing of discernment. Therefore, measuring conflict between the
the information has more changes, which is consistent with negative BPA and the initial BPA is also appropriate method
changes of entropy of random sets in Table 2. From the above, to denote the property of the negation. Let’s to calculate the
it can be proved that our proposed negation is reasonable. conflict between the initial BPA and its negation, using kxy to
In the following part, the increase of entropy of random indicate conflicts, x and y represents the xth or yth negation,
sets after taking negation was strictly show by mathematical for example, k01 represents the conflict between initial BPA

VOLUME 7, 2019 107011


TABLE 3. The trend of negation process of multiple elements.
and 1st negation, as follows. In kxy , x represents column, y
represents raw during Table. 3.
From the Table .3, it easily found that k01 is the lagest
during k0x , indicating that initial BPA and 1st negation has
maximum conflict. More interestingly, k0y can not always
keep increase from k01 to k09 , showing that conflict changes
with negation. Besides, there are more changes from k01 to
k03 , meaning that 1st negation and 2nd negation has a more
lager rate of consuming information, which is consistent with
the changes of entropy of random sets in Table .2. After taking
5th negation, the conflict has almost no change, meaning no
information consumption, meanwhile, entropy of random sets
has no change. In last, the BPA reaches convergence close to
uniform distribution and the conflict is no longer changing.
Therefore, measuring conflict between the negative BPA and
the initial BPA is also appropriate method to understand the
changes entropy of the BPAs.
V. APPLICATION OF PROPOSED METHOD
There are two sensors to recognize the target which maybe a,
b. The results from the sensors are as follows:
m1 (a) = 0.2, m1 (a, b) = 0.8
m2 (a) = 0.5, m2 (b) = 0.5
Using the Dempster’s rule of combination can get the follow-
ing result:
m (a) = 0.55, m (b) = 0.45
Using the proposed method to obtain a new information,
as follows:
m̄1 (a) = 0.4, m̄1 (b) = 0.5, m̄1 (a, b) = 0.1
m̄2 (a) = 0.25, m̄2 (b) = 0.25, m̄2 (a, b) = 0.4,
Similarly, using the Dempster’s rule of combination as fol-
lows:
m̄ (a) = 0.42, m̄ (b) = 0.52, m̄ (a, b) = 0.06
By comparing the results between the original and nega-
tion, the decision has more probability to regard target as a.
By analysing, the best target is is to have a larger original BPA
and smaller negation BPA. Hence, it can get two results from
the two sides based one data. This will improve the accuracy
of decision makers.
107012
X. Gao, Y. Deng: Negation of BPA
VI. CONCLUSION
This paper proposed a new method to obtain the negation
of the BPA and discussed some basic properties of proposed
negation method. Numerical examples are used to show the
changes of BPA taking negation. Conf Diss and NS are
used to measure changes of uncertainty during the nega-
tion process. Besides, the paper discussed the connection
between changes of uncertainty and entropy of random sets.
More importantly, the entropy of random sets always keep
increase after negation until it reaches its maximum, mean-
while, the BPA is uniform distribution. Moreover, conflict
plays an essential role in D-S theory, which is discussed
between initial BPA and its negation. It can be found that
the Conf , Diss, NS and conflict have the more changes in
the first negation than other negations, at the same time the
entropy of random sets also has the maximum change. From
this phenomenon, it can be known that consuming more
information means the changes of uncertainty with increase
of entropy. Summarizing, after several negation processes,
the BPA converges to uniform distribution with maximum
entropy of random sets without any changes.
CONFLICT OF INTERESTS
The authors declare that there is no conflict of interests
regarding the publication of this paper.
REFERENCES
[1] A. Janghorbani and M. H. Moradi, ‘‘Fuzzy evidential network and its appli-
cation as medical prognosis and diagnosis models,’’ J. Biomed. Inform.,
vol. 72, pp. 96–107, Aug. 2017.
[2] Z. Cao and C.-T. Lin, ‘‘Inherent fuzzy entropy for the improvement of
EEG complexity evaluation,’’ IEEE Trans. Fuzzy Syst., vol. 26, no. 2,
pp. 1032–1035, Apr. 2018.
[3] Y.-T. Liu, N. R. Pal, A. R. Marathe, and C.-T. Lin, ‘‘Weighted fuzzy
Dempster–Shafer framework for multimodal information integration,’’
IEEE Trans. Fuzzy Syst., vol. 26, no. 1, pp. 338–352, Feb. 2018.
[4] M. K. Ghorabaee, M. Amiri, E. K. Zavadskas, and J. Antuchevičienė,
‘‘Supplier evaluation and selection in fuzzy environments: A review of
MADM approaches,’’ Econ. Res.-Ekonomska Istraživanja, vol. 30, no. 1,
pp. 1073–1118, 2017.
[5] G. Shafer, A Mathematical Theory of Evidence, vol. 42. Princeton, NJ,
USA: Princeton Univ. Press, 1976.
[6] A. P. Dempster, ‘‘Upper and lower probabilities induced by a multivalued
mapping,’’ Ann. Math. Statist., vol. 38, no. 2, pp. 325–339, 1967.
[7] Z. Pawlak, ‘‘Rough sets,’’ Int. J. Comput. Inf. Sci., vol. 11, no. 5,
pp. 341–356, Oct. 1982.
[8] L. A. Zadeh, ‘‘A note on Z-numbers,’’ Inf. Sci., vol. 181, no. 14,
pp. 2923–2932, 2011.
[9] B. Kang, Y. Deng, K. Hewage, and R. Sadiq, ‘‘Generating Z-number based
on OWA weights using maximum entropy,’’ Int. J. Intell. Syst., vol. 33,
no. 8, pp. 1745–1755, 2018.
[10] H. Seiti, A. Hafezalkotob, S. E. Najafi, and M. Khalaj, ‘‘A risk-based fuzzy
evidential framework for FMEA analysis under uncertainty: An interval-
valued DS approach,’’ J. Intell. Fuzzy Syst., vol. 35, no. 2, pp. 1419–1430,
2018.
[11] H. Seiti, A. Hafezalkotob, and L. Martínez, ‘‘R-numbers, a new risk
modeling associated with fuzzy numbers and its application to decision
making,’’ Inf. Sci., vol. 483, pp. 206–231, May 2019.
[12] J. Abellán, C. J. Mantas, and J. G. Castellano, ‘‘A random forest approach
using imprecise probabilities,’’ Knowl.-Based Syst., vol. 134, pp. 72–84,
Oct. 2017.
[13] R. Wang, X. Gao, J. Gao, Z. Gao, and J. Kang, ‘‘An information transfer
based novel framework for fault root cause tracing of complex electrome-
chanical systems in the processing industry,’’ Mech. Syst. Signal Process.,
vol. 101, pp. 121–139, Feb. 2018.
VOLUME 7, 2019
X. Gao, Y. Deng: Negation of BPA
[14] Z. Zhang, D. Han, J. Dezert, and Y. Yang, ‘‘A new adaptive switching
median filter for impulse noise reduction with pre-detection based on
evidential reasoning,’’ Signal Process., vol. 147, pp. 173–189, Jun. 2018.
[15] J. Wang, K. Qiao, and Z. Zhang, ‘‘Trust evaluation based on evidence
theory in online social networks,’’ Int. J. Distrib. Sensor Netw., vol. 14,
no. 10, 2018, Art. no. 1550147718794629.
[16] A. D. Jaunzemis, M. J. Holzinger, M. W. Chan, and P. P. Shenoy, ‘‘Evidence
gathering for hypothesis resolution using judicial evidential reasoning,’’
Inf. Fusion, vol. 49, pp. 26–45, Sep. 2019.
[17] J. Vandoni, E. Aldea, and S. Le Hégarat-Mascle, ‘‘Evidential query-
by-committee active learning for pedestrian detection in high-density
crowds,’’ Int. J. Approx. Reasoning, vol. 104, pp. 166–184, Jan. 2019.
[18] Z. Wang, J.-M. Gao, R.-X. Wang, K. Chen, Z.-Y. Gao, and Y. Jiang, ‘‘Fail-
ure mode and effects analysis using Dempster–Shafer theory and TOPSIS
method: Application to the gas insulated metal enclosed transmission line
(GIL),’’ Appl. Soft Comput., vol. 70, pp. 633–647, Sep. 2018.
[19] J. Abellán and E. Bossé, ‘‘Drawbacks of uncertainty measures based on the
pignistic transformation,’’ IEEE Trans. Syst., Man, Cybern., Syst., vol. 48,
no. 3, pp. 382–388, Mar. 2018.
[20] B. Wu, X. Yan, Y. Wang, and C. G. Soares, ‘‘An evidential reasoning-based
CREAM to human reliability analysis in maritime accident process,’’ Risk
Anal., vol. 37, no. 10, pp. 1936–1957, 2017.
[21] M. Behrouz and S. Alimohammadi, ‘‘Uncertainty analysis of flood con-
trol measures including epistemic and aleatory uncertainties: Probability
theory and evidence theory,’’ J. Hydrologic Eng., vol. 23, no. 8, 2018,
Art. no. 04018033.
[22] Y. Li and Y. Deng, ‘‘Generalized ordered propositions fusion based
on belief entropy,’’ Int. J. Comput. Commun. Control, vol. 13, no. 5,
pp. 792–807, 2018.
[23] W. Zhu, Q. Ku, Y. Wu, H. Zhang, Y. Sun, and C. Zhang, ‘‘A research
into the evidence reasoning theory of two-dimensional framework and its
application,’’ Kybernetes, vol. 47, no. 5, pp. 873–887, 2018.
[24] S. Porebski and E. Straszecka, ‘‘Extracting easily interpreted diagnostic
rules,’’ Inf. Sci., vol. 426, pp. 19–37, Feb. 2018.
[25] Y. Han and Y. Deng, ‘‘An enhanced fuzzy evidential DEMATEL method
with its application to identify critical success factors,’’ Soft Comput.,
vol. 22, no. 15, pp. 5073–5090, 2018.
[26] R. R. Yager, ‘‘On the maximum entropy negation of a probability distribu-
tion,’’ IEEE Trans. Fuzzy Syst., vol. 23, no. 5, pp. 1899–1902, Oct. 2015.
[27] P. Smets, ‘‘The application of the matrix calculus to belief functions,’’ Int.
J. Approx. Reasoning, vol. 31, nos. 1–2, pp. 1–30, 2002.
[28] R. R. Yager, ‘‘On the measure of fuzziness and negation part I: Membership
in the unit interval,’’ Int. J. Gen. Syst., vol. 5, no. 4, pp. 221–229, 1979.
[29] N. Galatos and C. Tsinakis, ‘‘Generalized MV-algebras,’’ J. Algebra,
vol. 283, no. 1, pp. 254–291, 2005.
[30] H. Bustince, P. Burillo, and F. Soria, ‘‘Automorphisms, negations and
implication operators,’’ Fuzzy Sets Syst., vol. 134, pp. 209–229, Mar. 2003.
[31] K. C. Maes and B. De Baets, ‘‘Negation and affirmation: The role of
involutive negators,’’ Soft Comput., vol. 11, no. 7, pp. 647–654, 2007.
[32] D. E. Bell, ‘‘Regret in decision making under uncertainty,’’ Oper. Res.,
vol. 30, no. 5, pp. 961–981, 1982.
[33] L. Yin, X. Deng, and Y. Deng, ‘‘The negation of a basic probability assign-
ment,’’ IEEE Trans. Fuzzy Syst., vol. 27, no. 1, pp. 135–143, Jan. 2019.
[34] A. Srivastava and S. Maheshwari, ‘‘Some new properties of negation of a
probability distribution,’’ Int. J. Intell. Syst., vol. 33, no. 6, pp. 1133–1145,
2018.
[35] K. Xie and F. Xiao, ‘‘Negation of belief function based on the total
uncertainty measure,’’ Entropy, vol. 21, no. 1, p. 73, 2019.
[36] P. Dutta, ‘‘An uncertainty measure and fusion rule for conflict evidences of
big data via Dempster–Shafer theory,’’ Int. J. Image Data Fusion, vol. 9,
no. 2, pp. 152–169, 2018.
[37] M. Li, Q. Zhang, and Y. Deng, ‘‘Evidential identification of influen-
tial nodes in network of networks,’’ Chaos, Solitons Fractals, vol. 117,
pp. 283–296, Dec. 2018.
[38] L. Pan and Y. Deng, ‘‘A new belief entropy to measure uncertainty of
basic probability assignments based on belief function and plausibility
function,’’ Entropy, vol. 20, no. 11, p. 842, 2018.
[39] H. T. Nguyen, ‘‘On entropy of random sets and possibility distributions,’’
Anal. Fuzzy Inf., vol. 1, pp. 145–156, 1987.
[40] K. Chatterjee, D. Pamucar, and E. K. Zavadskas, ‘‘Evaluating the perfor-
mance of suppliers based on using the R’AMATEL-MAIRCA method for
green supply chain implementation in electronics industry,’’ J. Cleaner
Prod., vol. 184, pp. 101–129, May 2018.
VOLUME 7, 2019
[41] H. Yang, Y. Deng, and J. Jones, ‘‘Network division method based on
cellular growth and physarum-inspired network adaptation,’’ Int. J. Uncon-
ventional Comput., vol. 13, no. 6, pp. 477–491, 2018.
[42] P. Xu, R. Zhang, and Y. Deng, ‘‘A novel visibility graph transformation of
time series into weighted networks,’’ Chaos, Solitons Fractals, vol. 117,
pp. 201–208, Dec. 2018.
[43] L. Fei and Y. Deng, ‘‘A new divergence measure for basic probability
assignment and its applications in extremely uncertain environments,’’ Int.
J. Intell. Syst., vol. 34, no. 4, pp. 584–600, 2019.
[44] H. Seiti and A. Hafezalkotob, ‘‘Developing pessimistic-optimistic risk-
based methods for multi-sensor fusion: An interval-valued evidence theory
approach,’’ Appl. Soft Comput., vol. 72, pp. 609–623, Nov. 2018.
[45] J.-B. Yang and D.-L. Xu, ‘‘Evidential reasoning rule for evidence combi-
nation,’’ Artif. Intell., vol. 205, pp. 1–29, Dec. 2013.
[46] F. Xiao, ‘‘A hybrid fuzzy soft sets decision making method in medical
diagnosis,’’ IEEE Access, vol. 6, pp. 25300–25312, 2018.
[47] Z. He and W. Jiang, ‘‘An evidential dynamical model to predict the inter-
ference effect of categorization on decision making,’’ Knowl.-Based Syst.,
vol. 150, pp. 139–149, Jun. 2018.
[48] D. Sani, H. Tutut, A. L. Maizatul, C. Haruna, I. A. Adamu, and
M. Z. Akram, ‘‘A review on soft set-based parameter reduction and deci-
sion making,’’ IEEE Access, vol. 5, pp. 4671–4689, 2017.
[49] Z. He and W. Jiang, ‘‘An evidential Markov decision making model,’’ Inf.
Sci., vol. 467, pp. 357–372, Oct. 2018.
[50] Y. Zhang, Y. Liu, Z. Zhang, and N. Zhao, ‘‘Collaborative fusion for
distributed target classification using evidence theory in IoT environment,’’
IEEE Access, vol. 6, pp. 62314–62323, 2018.
[51] Z. Liu, Q. Pan, J. Dezert, J.-W. Han, and Y. He, ‘‘Classifier fusion with
contextual reliability evaluation,’’ IEEE Trans. Cybern., vol. 48, no. 5,
pp. 1605–1618, May 2018.
[52] Z.-G. Su, T. Denoeux, Y.-S. Hao, and M. Zhao, ‘‘Evidential K-NN classi-
fication with enhanced performance via optimizing a class of parametric
conjunctive t-rules,’’ Knowl.-Based Syst., vol. 142, pp. 7–16, Feb. 2018.
[53] R. Jiroušek and P. P. Shenoy, ‘‘A new definition of entropy of belief
functions in the Dempster–Shafer theory,’’ Internal J. Approx. Reasoning,
vol. 92, pp. 49–65, Jan. 2018.
[54] L. Yin and Y. Deng, ‘‘Toward uncertainty of weighted networks:
An entropy-based model,’’ Phys. A, Stat. Mech. Appl., vol. 508,
pp. 176–186, Oct. 2018.
[55] X. Deng, W. Jiang, and Z. Wang, ‘‘Zero-sum polymatrix games with link
uncertainty: A Dempster–Shafer theory solution,’’ Appl. Math. Comput.,
vol. 340, pp. 101–112, Jan. 2019.
[56] G. Sun, X. Guan, X. Yi, and J. Zhao, ‘‘Conflict evidence measurement
based on the weighted separate union kernel correlation coefficient,’’ IEEE
Access, vol. 6, pp. 30458–30472, 2018.
[57] J.-B. Yang and D.-L. Xu, ‘‘A study on generalising Bayesian inference
to evidential reasoning,’’ in Proc. Int. Conf. Belief Functions, 2014,
pp. 180–189.
[58] I. Friedberg, X. Hong, K. Mclaughlin, P. Smith, and P. C. Miller, ‘‘Eviden-
tial network modeling for cyber-physical system state inference,’’ IEEE
Access, vol. 5, pp. 17149–17164, 2017.
[59] F. Xiao, ‘‘Multi-sensor data fusion based on the belief divergence measure
of evidences and the belief entropy,’’ Inf. Fusion, vol. 46, pp. 23–32,
Mar. 2019.
[60] Y. Zhou, X. Tao, L. Luan, and Z. Wang, ‘‘Safety justification of train move-
ment dynamic processes using evidence theory and reference models,’’
Knowl.-Based Syst., vol. 139, pp. 78–88, Jan. 2018.
[61] Y. Deng, ‘‘Generalized evidence theory,’’ Appl. Intell., vol. 43, no. 3,
pp. 530–543, Oct. 2015.
[62] R. Sun and Y. Deng, ‘‘A new method to identify incomplete frame of
discernment in evidence theory,’’ IEEE Access, vol. 7, pp. 15547–15555,
2019.
[63] H. Mo and Y. Deng, ‘‘A new MADA methodology based on D numbers,’’
Int. J. Fuzzy Syst., vol. 20, no. 8, pp. 2458–2469, 2018.
[64] F. Xiao, ‘‘A novel multi-criteria decision making method for assessing
health-care waste treatment technologies based on D numbers,’’ Eng. Appl.
Artif. Intell., vol. 71, pp. 216–225, May 2018.
[65] X. Deng and W. Jiang, ‘‘An evidential axiomatic design approach for
decision making using the evaluation of belief structure satisfaction to
uncertain target values,’’ Int. J. Intell. Syst., vol. 33, no. 1, pp. 15–32, 2018.
[66] Z. Liu, Q. Pan, J. Dezert, and A. Martin, ‘‘Combination of classifiers with
optimal weight based on evidential reasoning,’’ IEEE Trans. Fuzzy Syst.,
vol. 26, no. 3, pp. 1217–1230, Jun. 2018.
107013
 X. Gao, Y. Deng: Negation of BPA

[67] W. Jiang, ‘‘A correlation coefficient for belief functions,’’ Int. J. Approx. XIAOZHUAN GAO is currently pursuing the
Reasoning, vol. 103, pp. 94–106, Dec. 2018. Ph.D. degree with the Institute of Fundamental
[68] Y. Song, X. Wang, L. Lei, and S. Yue, ‘‘Uncertainty measure for interval- and Frontier Science, University of Electronic Sci-
valued belief structures,’’ Measurement, vol. 80, pp. 241–250, Feb. 2016. ence and Technology of China, Chengdu, China.
[69] G. Shafer, ‘‘Dempster’s rule of combination,’’ Int. J. Approx. Reasoning, Her research interests include evidence theory,
vol. 79, pp. 26–40, Dec. 2016. decision making, information fusion, and quantum
[70] X. Su, L. Li, F. Shi, and H. Qian, ‘‘Research on the fusion of depen- computation.
dent evidence based on mutual information,’’ IEEE Access, vol. 6,
pp. 71839–71845, 2018.
[71] J. Wang, K. Qiao, and Z. Zhang, ‘‘An improvement for combination rule in
evidence theory,’’ Future Gener. Comput. Syst., vol. 91, pp. 1–9, Feb. 2019.
[72] Y. Deng, ‘‘Deng entropy,’’ Chaos, Solitons Fractals, vol. 91, pp. 549–553,
Oct. 2016.
[73] S. C. Jugade and A. C. Victorino, ‘‘Grid based estimation of decision
uncertainty of autonomous driving systems using belief function theory,’’
IFAC-PapersOnLine, vol. 51, no. 9, pp. 261–266, 2018.
[74] J. Abellán, ‘‘Analyzing properties of Deng entropy in the theory of evi-
dence,’’ Chaos, Solitons Fractals, vol. 95, pp. 195–199, Feb. 2017.
[75] Y. Yang and D. Han, ‘‘A new distance-based total uncertainty measure in
the theory of belief functions,’’ Knowl.-Based Syst., vol. 94, pp. 114–123,
Feb. 2016.
[76] P. Dutta, ‘‘Modeling of variability and uncertainty in human health risk
assessment,’’ MethodsX, vol. 4, pp. 76–85, 2017.
[77] R. R. Yager, ‘‘Entropy and specificity in a mathematical theory of evi-
dence,’’ Int. J. Gen. Syst., vol. 9, no. 4, pp. 249–260, 1983.
[78] U. Höhle, ‘‘A general theory of fuzzy plausibility measures,’’ J. Math. Anal. YONG DENG received the Ph.D. degree in precise
Appl., vol. 127, no. 2, pp. 346–364, 1987.
instrumentation from Shanghai Jiao Tong Uni-
[79] R. V. L. Hartley, ‘‘Transmission of information,’’ Bell Syst. Tech. J., vol. 7,
versity, Shanghai, China, in 2003. From 2005 to
no. 3, pp. 535–563, 1928.
[80] D. Dubois and H. Prade, ‘‘A note on measures of specificity for fuzzy sets,’’ 2011, he was an Associate Professor with the
Int. J. Gen. Syst., vol. 10, no. 4, pp. 279–283, 1985. Department of Instrument Science and Technol-
[81] X. Deng, ‘‘Analyzing the monotonicity of belief interval based uncertainty ogy, Shanghai Jiao Tong University. In 2010, he
measures in belief function theory,’’ Int. J. Intell. Syst., vol. 33, no. 9, was a Professor with the School of Computer
pp. 1869–1879, Sep. 2018. and Information Science, Southwest University,
[82] L. Chen, L. Diao, and J. Sang, ‘‘Weighted evidence combination rule based Chongqing, China. In 2012, he was an Visiting
on evidence distance and uncertainty measure: An application in fault Professor with Vanderbilt University, Nashville,
diagnosis,’’ Math. Problems Eng., vol. 2018, 2018, Art. no. 5858272. TN, USA. In 2016, he was also a Professor with the School of Electronic
[83] R. Clausius, The Mechanical Theory of Heat: With Its Applications to the and Information Engineering, Xi’an Jiaotong University, Xian, China. Since
Steam-Engine and to the Physical Properties of Bodies. London, U.K.: 2017, he has been a Full Professor with the Institute of Fundamental and
J. van Voorst, 1867. Frontier Science, University of Electronic Science and Technology of China,
[84] P. Liu and X. Zhang, ‘‘Approach to multi-attributes decision making with Chengdu, China. He has published more than 100 papers in refereed jour-
intuitionistic linguistic information based on Dempster–Shafer evidence nals, such as Decision Support Systems, European Journal of Operational
theory,’’ IEEE Access, vol. 6, pp. 52969–52981, 2018.
Research, and Scientific Reports. His research interests include evidence
[85] A. Rajan, Y. C. Kuang, M. P.-L. Ooi, S. N. Demidenko, and H. Carstens,
theory, decision making, information fusion, and complex system modeling.
‘‘Moment-constrained maximum entropy method for expanded uncer-
tainty evaluation,’’ IEEE Access, vol. 6, pp. 4072–4082, 2018. He received numerous honors and awards, including the Elsevier Highly
[86] Y. Tang, D. Zhou, and F. T. S. Chan, ‘‘An extension to Deng’s entropy in the Cited Scientist in China, from 2014 to 2018. He served as the program
open world assumption with an application in sensor data fusion,’’ Sensors, member of many conferences, such as the International Conference on Belief
vol. 18, no. 6, p. 1902, 2018. Functions. He has served as a member of many editorial board, such as
[87] F. Xiao, ‘‘An improved method for combining conflicting evidences based Academic Editor of the PLOS ONE. He served as the Reviewer for more
on the similarity measure and belief function entropy,’’ Int. J. Fuzzy Syst., than 30 journals, such as the IEEE TRANSACTION ON FUZZY SYSTEMS.
vol. 20, no. 4, pp. 1256–1266, 2018.

107014 VOLUME 7, 2019