An Improved Evidence Combination Method of D-S Theory

Dempster-Shafer evidence theory provides a useful computational scheme for integrating uncertain information from multiple sources. However, the combination result can be paradoxical when the evidences seriously conflict with each other. In this paper, we present an improved D-S algorithm which verifies the weights of conflicting evidences. Firstly, we calculate the distance between any two of the evidences by the Jousselme distance formula and obtain the entropy index of distance correspondingly by using information entropy theory. Secondly, compute the relative support degree and absolute support degree so as to modifies the basic probability assignment (BPA). Experiments show that this method has a more comprehensive consideration of conflict evidences problem and more sensitive on the differences of focal element.