An Improved Conflicting Evidence Fusion Method in D-S Theory

Abstract

To solve the invalidation problem of Dempster-Shafer evidence theory for high conflicting fusion, this paper improves D-S algorithm based on weighted average method. A new similarity function is proposed to represent similarity, and by normalized similarity we obtain support to pretreated evidences, then, fuse preprocessed evidences by Dempster's rule. The more expected result data shows that, compared with other methods, this new algorithm can reduce harmful influence of false evidence effectively, at the same time, have higher convergence rate and reduce decision risk. Finally, the new similarity function is analyzed to prove the rationality of improved algorithm.