The Improvement And Application Of Dempster Rule Of Combination

Uncertainty reasoning is used to deal with uncertain and incomplete information.D-S evidence theory is an important method of uncertain reasoning.Other methods for dealing with uncertain reasoning include credibility methods,Bayesian inference,fuzzy integral methods,rough set theory,etc.Each of these methods has its own advantages.The unique advantage of the evidence theory is that it can combine prior probabilities like Bayesian inference,handle the concept of fuzzy sets like languages and has a wide range of applications.Dempster's rule of combination is an important part of the D-S evidence theory.Dempster's rule of combination is widely used in expert consultation systems,medical diagnosis,risk decision-making,image recognition and so on.However,the original fusion rule has its inherent drawbacks due to its normality.Many scholars have put forward many solutions to the problem of the inability to deal with conflicting evidence due to normality.Some methods have relatively good performance under certain conditions.However,many of them have some inherent disadvantages.It is difficult to achieve a balance between property,convergence,and computational complexity.This paper summarizes the unified form of distance between the evidence and uses new expressions to illustrate that the conflict coefficient k is not a suitable criterion for judging conflicts.This paper consider the two-dimensional distance and use the method of defining new distances between the evidence.We analyze the basic probability assignment of each evidence to analyze the degree of similarity between the evidence.We sort the evidence by importance according to the size of support to select the important evidence and general evidence.The paper assigns weight to each piece of evidence so as to achieve the purpose of eliminating the degree of conflict between the evidence.Then we use Dempster's rule of combination to calculate the result to get relatively reasonable results.Then the superiority of the method is proved through actual examples.Finally,this paper applies the new probabilistic transformation method to the combined results or a single BPA,and combines the risk decision knowledge to solve related problems.