Skip to Main Content
A new method based on multi-objective genetic algorithm (MOGA) is proposed to extract parsimonious fuzzy rule bases for modeling nonlinear MIMO dynamical systems. Structure selection, parameter estimation, model performance and model validation are important objectives in the process of non-linear system identification. MOGA is applied to these multiple, conflicting objectives and yields a set of candidate parsimonious and valid fuzzy models. The algorithm combines the advantages of genetic algorithms strong search capacity and recursive least square (RLS) fast convergence. Antecedent parts of a complete fuzzy model and inclusion/exclusion of fuzzy rules are coded into a chromosome. Then RLS is used to determine the consequent parts of selected rules. The practical applicability of the proposed algorithm is examined by an industrial nonlinear system modeling benchmark problem.