Clustering belief functions based on attracting and conflicting metalevel evidence

Abstract

This chapter focuses on the method for clustering belief functions based on attracting and conflicting metalevel evidence. Such clustering is done when the belief functions concern multiple events, and all belief functions are mixed up. The chapter extends an earlier method within Dempster-Shafer theory for handling belief functions that concern multiple events. The belief functions are clustered into subsets that should be handled independently. These conflicts were interpreted as metalevel evidence about the partition of the set of belief functions. Each piece of conflicting metalevel evidence states that the two belief functions do not belong to the same subset. The previously developed method is extended into also being able to handle the case of attracting metalevel evidence. Such evidence is not generated internally in the same way as the conflicting metalevel evidence. Instead, it is assumed that it is given from some external source as additional information about the partitioning of the set of all belief functions.