Understanding the Inference Mechanism of FURIA by means of Fingrams

Resumo

This paper shows the use of Fingrams –Fuzzy Inference-grams– aimed at unveiling graphically some hidden details in the usual behavior of the precise fuzzy modeling algorithm FURIA –Fuzzy Unordered Rule Induction Algorithm–. FURIA is recognized as one of the most outstanding fuzzy rule-based classification methods attending to accuracy. Although FURIA usually produces compact rule bases, with low number of rules and antecedents per rule, its interpretability is arguable, being penalized by the absence of linguistic readability and a complex inference mechanism. Fingrams offer a methodology for visual representation and exploratory analysis of fuzzy rule-based systems. FURIA-Fingrams, i.e. fuzzy inference-grams representing fuzzy systems learnt with FURIA, make easier understanding the FURIA inference mechanism thanks to the possibilities they offer: detecting instances not covered by any rule; highlighting important rules; clarifying the so-called stretching mechanism; etc.