ACPOP: Ambiguity correction-based pseudo-outer-product fuzzy rule identification algorithm

Fuzzy rules generated from neuro-fuzzy systems may contain ambiguous rules, due to numerous factors. While contradiction-correction often ensures consistency in fuzzy rule-bases, a differing approach should be reserved for problems where the linguistic definitions can be mutually-inclusive. For these cases, the proposed ambiguity-correction approach is a simple procedure that prevents excessive skew towards stronger rules, and still creates consistent fuzzy rule-base. This paper describes a proof-of-concept model, ACPOP-CRI(S), where ambiguity-correction can be adapted to the generic POP-CRI(S) framework. Experimental results on the Nakanishi dataset shows that the ACPOP rule identification algorithm has the potential to perform better, and generate fewer rules than the generic POP algorithm.