A disscussion on the overlapping in Fuzzy Production rule reasoning

When the Fuzzy Production Rules (FPRs) are used for knowledge representation in rule-based system, overlapping exists among these rule sets corresponding to different classes. Each of the rule sets is composed of the rules which have the same consequence. In order to model and handle this kind of overlapping, this paper proposes a nonnegative non-additive set function and provides a new algorithm with respect to the new set function for rule-based system based on possibility measure and necessity measure, instead of assigning maximum similarity measure of sets of FPRs as the reasoning result. Experimental results on a number of selected databases show that handling overlapping in FPRs in this way can improve the reasoning accuracy and strengthen the reasoning capability of rule-based system.