An information fusion approach based on weight correction and evidence theory

Abstract

The combination rules of the Dempster-Shafer evidence theory can lead to illogical results for highly conflicting evidence from different information sources. We propose a general formula for conflict degree calculation from the perspective of modifying evidence sources as a weighted sum of conflict coefficients and Jousselme distance, and the new metric of focal element dispersion is defined by adaptively adjusting the ratio of these two metrics: if the focal element dispersion is too high, the impact of conflict coefficients is increased, and vice versa. We then define the concept of preference consistency and propose a formula for calculating this metric that redistributes the weights of individual pieces of evidence based on the preferences of all evidence. Finally, typical examples show that the proposed rules can manage conflicting evidence with better convergence and interference resistance.