The Study Of Conflict Evidence In D-S Evidence Theory And Its Application

D-S evidence theory has great flexibility in dealing with uncertainty information and has a wide range of applications.However,the D-S evidence theory is often faced with evidence of inconsistency or conflict in practical application,and the quantitative measurement and synthesis of these conflicting evidence aroused a great deal of scholars' research.In this paper,the D-S evidence theory is used to measure the conflict of evidence,with which combine the conflict of evidence and construct the basic probability assignment function,the main research work is as follows:(1)Aiming at the open issue that the classical conflict coefficient and Jousselme distance measure can not correctly judge the conflict evidence,a new measurement of conflict was presented based on the definition of similarity coefficient and Jaccard matrix.Theoretic proof and numreical examples verify the efficiency of the proposed measurement.A typical example is analyzed and compared with other methods.It has been shown by comparison with other methods that the proposed measurement can effectively measure the degree of conflict between evidences.(2)To remedy the shortcomings in the recognition of evidence conflict by using conventional conflict coefficient,this method converted the Pignistic probability distance to the credibility of evidence.Then it combined credibility and false degree,which to construct a new correction coefficient to discount the basic probability assignment function.The basic probability assignment functions adjusted by correction coefficient are fused by D-S combination rule.Numerical example analysis results shows that the improved evidence combination method have the possession of rapid convergence and good robustness,keeping the good nature of D-S combination rules.(3)One of the open issues of D-S theory is how to determine basic probability assignment function.However,there is still no uniform standard to determine basic probability assignment function.A method to determine basic probability assignment function based on interval distance is proposed in the paper.At first,the model of interval distance is constructed with the samples.Then the interval diatance is normalized to get the value of basic probability assignment function.The effectiveness of this method is proved by classifying the Iris Data.