Dempster-Shafer이론을 이용한 징후 결합 방법

The Dempster-Shafer theory offers a mathematically solid framework for fusing evidence. According to the theory, the belief on a hypothesis is computed by aggregating all the basic belief, called bpa, committed to itself and its subsets. However, since the set of subsets grows exponentially with the number of elements, not all subsets can be counted for evidence combination. Therefore, various approximation schemes are proposed to restrict the subsets being considered. In this paper, a new approximation scheme is proposed which restricts the subsets as the given hypotheses and their complements. That is, if a new hypothesis generated by the Dempster's orthogonal sum is not contained in the given hypotheses and their complements, its belief is assigned to the nearest supersets among the hypotheses and their complements. The proposed scheme yields much closer approximation and still maintains the concept of the belief interval. The performance of the proposed scheme isS verified through an evaluation with a sample data set.