Abstract:

There are some defects with Dempster’s rule of combination. It may draw a counter-intuitive conclusion when combining highly conflicting evidences and cannot deal with contradictory evidences. Moreover, it cannot converge according to the cardinality of the sets in the evidences. An evidence fusion method is proposed which is an integration of two different conflict treatment strategies, i.e., modifying the combination rule and modifying the evidences. The original evidences are modified by a weighted average method based on Jousselme’s measure of distance between evidences and a new combination rule is proposed to combine the modified evidences for several times. The comparative study with current several representative methods show that the proposed method performs well in converging according to the cardinality of sets, convergence trend and dealing with conflicting evidences.