AN IMPROVEMENT TO KOCZY AND HIROTAS INTERPOLATIVE REASONING IN SPARSEFUZZY RULE BASES

In sparse fuzzy rule bases, conventional fuzzy reasoning methods canno t reach a proper conclusion. To tackle this problem, Koczy and Hirota have proposed a method called interpolative reasoning. It has been fou nd that by this method the convexity of the reasoning consequence fuzz y set cannot always be retained. In this paper, the authors give a gen eral convex condition for Koczy and Hirota's method and, starting from this condition, propose an improvement to the method. Firstly, from t he given rules in the sparse rule base is constructed a new rule which is near to the antecedent fuzzy set. Then the reasoning is performed with this new rule, based on similarities of fuzzy sets in the anteced ent and consequent parts. It is shown that the proposed method maintai ns the logical interpretation of modus ponens and guarantees the norma lity and convexity of the reasoning consequence fuzzy set in some clas ses of fuzzy rules. (C) 1996 Elsevier Science Inc.