Identification of recurrent fuzzy systems with genetic algorithms

Abstract

This work presents an algorithm for identification of fuzzy recurrent models of non-linear dynamic systems. The identification algorithm is based on a general purpose genetic algorithm. The resulting recurrent fuzzy system can encode into a fuzzy finite state automaton in which the linguistic terms of the fuzzy model are the states, and rule base weights are transition possibilities. The identification algorithm is tested against benchmark identification problems found in the literature.