A conflict evidence fusion method based on the composite discount factor and the game theory

Abstract

Dempster–Shafer (D–S) evidence theory is widely used in various fields of information fusion. However, it is still an open issue that the D–S evidence theory may produce the counter–intuitive results in fusing high–conflict evidences. Aim at this problem, a novel conflict evidence fusion method based on the composite discount factor and the game theory is proposed in this paper. Firstly, an improved Shafer's conflict measurement formula based on the Jaccard similarity coefficient is devised, and combined with the Jousselme distance into a novel binary function to measure the global conflict between evidences as the evidence falsity. Then, the local conflict between evidences and the information volume of evidences are measured by using the Jousselme distance and belief entropy to indicate the credibility and uncertainty of evidences. Next, based on the game theory, the falsity, credibility and uncertainty are weighted and combined into the composite discount factors to correct each body of evidence (BOE). Ultimately, all corrected evidences are fused by Dempster's combination rule to obtain the final result. Two numerical examples are given to verify that the proposed method is effective and feasible, which outperforms the previous methods in handling the conflict evidences.