An Improved Evidence Theory Based on Random Set

D-S evidence theory is an effective tool to deal with uncertain information fusion, but there is a disadvantage in dealing with high confliction between evidences. Considering different evidences have different reliabilities, this paper, which is based on the framework of random set, proposed a concept of relative information entropy to acquire the reliability weight of evidence itself and evidence distance based on belief function to acquire the reliability weight of the relationship between evidences. Thus, numerical experiments are adopted to prove its effectiveness.