ADAPTIVE FUZZY RULE-BASED CLASSIFICATION SYSTEMS

This paper proposes an adaptive method to construct a fuzzy rule-based classification system with high performance for pattern classificatio n problems. The proposed method consists of two procedures: an error c orrection-based learning procedure and an additional learning procedur e. The error correction-based learning procedure adjusts the grade of certainty of each fuzzy rule by its classification performance, That i s, when a pattern is misclassified by a particular fuzzy rule, the gra de of certainty of that rule is decreased. On the contrary, when a pat tern is correctly classified, the grade of certainty is increased, Bec ause the error correction-based learning procedure is not meaningful a fter all the given patterns are correctly classified, we cannot adjust a classification boundary in such a case,To acquire a more intuitivel y acceptable boundary, we propose an additional learning procedure. We also propose a method for selecting significant fuzzy rules by prunin g unnecessary fuzzy rules, which consists of the error correction-base d learning procedure and the concept of forgetting. We can construct a compact fuzzy rule-based classification system with high performance, Finally, we test the performance of the proposed two methods on the w ell-known iris data.