Two Efficient Combination Rules for Conflicting Belief Functions

According to the framework of Dempster-Shafer evidence theory, information fusion relies on the use of a combination rule allowing the belief functions for the different propositions to be combined. Dempster´s rule of combination is a basic rule of combination. However, Dempster´s combination operator is poor in the management of the conflict among the various information sources at the normalization step. In this paper, different importance of each body of evidence to be combined is considered, and the distance or the conflicting degree between two bodies of evidence is used in determining the importance of evidence. Based on two different measures of relative importance of evidence, we define two weighting schemes for the average support degrees of the propositions. In the two proposed combination rules, the conflicting mass is assigned to propositions according to the weighted average support degrees instead of normalization. Experiments show that the two proposed combination rules can efficiently handle conflicting evidences, and improve the reliability and rationality of the combination results compared with Dempster´s rule and other alternatives.