A Novel Combination Method for Conflicting Evidence in Open Frame of Discernment

Abstract

Dempster-Shafer theory is a useful method in data fusion. To avoid counter-intuitive result of classical Dempster-Shafer's combination rule dealing with highly conflict evidences, many improved combination methods have been proposed through modifying the classical Dempster-Shafer's rule or original evidence by using a certain measure of the degree of conflict or uncertain data. However, usually the frame of discernment is incomplete, the current methods have limitation in handling this condition.In order to solve this problem, an improved combination method for conflicting evidence in open frame of discernment is proposed. In this paper, according to information theory the information entropy of each evidence is calculated and assigned to the unknown proposition which is added into the frame of discernment. Then reliability degree of evidences obtained by the Jousselme distance is used to revise source evidences. Finally, Dempster-Shafer's rule is applied to combine the mean of revised evidences several times. One numerical example is employed and in comparison with the existing alternatives, the proposed approach can effectively solve the conflicting problem and the incompleteness of frame.