Fuzzy modeling with hybrid systems

Abstract

In this paper we present different approaches to the problem of fuzzy rules extraction by using a combination of fuzzy clustering and genetic algorithms as the main tools. This combination of techniques let us define a hybrid system by which we can have different approaches in a fuzzy modeling process. For example, we can obtain a first approximation to the fuzzy rules that describe the system behavior represented by a collection of raw data, without any assumption about the structure of the data using a fuzzy clustering technique, and subsequently, these rules can be tuned using a genetic algorithm. Alternatively, this genetic algorithm can be used in order to generate and tune the fuzzy rules directly from the data with or without some priori information. Finally, their performances are compared.