Generating Basic Probability Assignment from the View of Distance Measures and Its Application in Evidential Decision Tree

In Dempster-Shafer evidence theory, basic probability assignment (BPA) plays a important role in representing uncertain and unknown information. How to generate highquality BPA is essential, which can promote further application of evidence theory. BPA can be understood as the possibility assigned to each proposition. Distance measures, as an effective tool for quantitatively analyzing inconsistencies between samples and known information, are central to this process. Hence, this paper proposes a novel method to generate BPA by comprehensively considering Gaussian distribution and distance measures. New method is applied to evidential decision tree and its effectiveness can be verified by real world data set.