A new base basic probability assignment approach for conflict data fusion in the evidence theory

Dempster-Shafer evidence theory (D-S theory) is applied to process uncertain information in different scenarios. However, traditional Dempster combination rule may produce counterintuitive results while dealing with highly conflicting data. Inspired by a perspective of constructing base belief function for conflicting data processing in D-S theory, a new base basic probability assignment (bBPA) method is proposed to process the potential conflict before data fusion. Instead of assigning initial belief on the whole power set space, the new method assigns the base belief to basic events in the frame of discernment. Consequently, the bBPA is consistent with the classical probability theory. Several numerical examples are adopted to verify the reliability and accuracy of the method in processing highly conflicting data. The data sets in the University of California Irvine (UCI) Machine Learning Repository are used to verity the availability of the new method in classification problem. Experimental result shows that the new method has some superiority in dealing with highly conflicting data.