Acquiring Knowledge from Decision Tables for Evidential Reasoning

This paper proposes a method to acquire rules for evidential reasoning from multiple decision tables. The knowledge acquisition consists of two steps: the first step derives uncertain rules of the form: if a(u) is x, then d(u) is in Y(1) with r(1) or ... or d(u) is in Y(M) with r(M), where r(j)(j = 1, ..., M) are beliefs represented by basic belief assignment. The second step derives rules of the form: if a(u) is in X then d(u) is in Y(1) with r(1) or ... or d(u) is in Y(M) with r(M), from the rules obtained in the first step. A non-specificity based condition imposed on rules generated in the second step is introduced. It is also shown that the disjunctive combination approach satisfies the condition.