Novel approach for measuring the conflict between evidence

Tian-tian BAO,Xin-lian XIE,Zhao-kun WEI
DOI: https://doi.org/10.7641/CTA.2017.60275
2017-01-01
Abstract:To solve the invalidation problems when the evidence theory is used to deal with the cases with highly conflicting evidence, a new method is investigated for measuring and representing the degree of conflict between evidence. First of all, based on the comparison and analysis of some existing evidence distance methods which are the main conflict measures, a new probabilistic-based distance is defined and combined with Jousselme's distance to describe the difference between evidence. Next, because the conception of evidence distance cannot substitute for the definition of evidential conflict completely, a cross entropy of basic probability assignments (BPAs) is proposed based on the idea of fuzzy cross entropy to present the degree of divergence between evidence. Then, in the light of mixing the distance and divergence between evidence by applying Hamacher T-conorm, a new method for measuring the degree of conflict between evidence is put forward. Finally, the numerical examples verify the efficiency of the proposed method.
What problem does this paper attempt to address?