Researches On Degree Of Reliability In Conflicting Evidence Fusion

In recent years,as society continues to accelerate the process of information,the human have entered the era of all things interconnected,how to effectively integrate different levels of multi-source information has become the focus of current national government departments and research institutions.As one of the most active research hotspots,information fusion technology has been applied effectively in military field,with the development of artificial intelligence and Internet of things,information fusion has been widely concerned in people's daily life.People are enjoying the information fusion technology to bring convenience to life while taking the risk of technical defects,such as miscarriage of medical decision making,confusion of information processing,etc.The processing of multi-source information is mainly about the analysis and modeling of information.Currently,the most commonly used methods of uncertain information modeling are probability theory and mathematical statistics-based method,Dempster-Shafer evidence theory method,fuzzy set theory and rough set theory.Each method has its own advantages and disadvantages,the advantage of Dempster-Shafer evidence theory in the representation of stochastic uncertainty and subjective uncertainty,the ability to carry out uncertain reasoning in the absence of a priori information and many other excellent features attracting wide concern among researchers.Although Dempster-Shafer evidence theory has many advantages,there are still some problems to be improved in practice.For instance,the generation of basic belief assignment,the processing of dependent evidence and the fusion of conflicting evidence,etc.Among them,how to effectively deal with conflicting evidence has been the focus of research personnel of all over the world.This paper mainly analyzes the reliability of the evidence in the conflicting evidence fusion and propose two methods to deal with the conflict of evidence.Firstly,the information entropy is introduced into the uncertainty evaluation of evidence,the information entropy instead the traditional method to evaluate the reliability of different evidence,and using the improved method to carry out the weighted evidence fusion.Then,a reliability evaluation method based on belief interval is proposed,in proposed method,the reliability of the evidence is evaluated by using the disagreement of the belief intervals of different evidence,and two different distance measures are used to fuse the evidence.In order to verify the effectiveness and practicability of the proposed method,we compare several conflicting management methods on the same data,the effectiveness of the new method is verified.The details are as follows:(1)Proposing of a method for conflict management based on information entropy and fuzzy preference relationsThe existing studies have primarily focused on investigating similarity of evidence.However,the similarity measurement is highly dependent on the capability of distance functions and will substantially increase the computational complexity.So,we propose a new method based on the variance of information entropy to evaluate the disagreement between evidence,we use fuzzy preference relations to obtain the reliability of different evidence.The obtained evidence reliability is used to modify the original evidence and then to fuse the evidence with Dempster's rule of combination,which can effectively solve the conflicting evidence fusion problem.(2)Proposing of the reliability evaluation of evidence based on belief intervalsDistance measure is adopted by most researchers in conflicting evidence management,The effectiveness of evidence distance used in conflict management become the crucial element of this approach.Therefore,we use the distance measure based on belief intervals to evaluate the reliability of evidence.We apply the Euclidean-family belief interval-based distance and Chebyshev-family belief interval-based distance as the measurement,the reliability of different evidences is obtained by the method of disagreement of evidence.Because the greater the interval distance between different bodies of evidence,the greater the difference between the evidence,then the principle of majority is used to find the outliers in multiple evidence,and assign the outlier low credibility before evidence fusion.