COMBINATION METHODS AND CONFLICT HANDLING IN EVIDENTIAL THEORIES

Abstract

Evidence theory has been acknowledged as an important approach to dealing with uncertain, incomplete and imperfect information. In this framework, different formal techniques have been developed in order to address information aggregation and conflict handling. The variety of proposed models clearly demonstrates the range of possible underlying assumptions in combination rules. In this paper we present a review of some of the most important methods of combination and conflict handling in order to introduce a more generic rule for aggregation of uncertain evidence. We claim that the models based on mass multiplication can address the problem domains where randomness and stochastic independence is the dominant characteristic of information sources, although these assumptions are not always adhered to many practical cases. The proposed combination rule here is not only capable of retrieving other classical models, but also enables us to define new families of aggregation rules with more flexibility on dependency and normalization assumptions.