A Multi-Objective Cooperative Coevolutionary Approach to Mamdani Fuzzy System Generation

Alessio Botta, Pietro Ducange, Beatrice Lazzerini, Francesco Marcelloni.

A novel multi-objective cooperative coevolutionary approach aimed at generating a set of Mamdani-type fuzzy rule-based systems (FRBSs) with optimal trade-offs between accuracy and interpretability is proposed. Interpretability is measured both in terms of complexity of the rule base (RB) and of integrity of the data base (DB). In the framework of the cooperative coevolutionary approach, multi-objective optimization of RB and DB is performed in two distinct populations. Individuals of the two populations cooperate among them through representatives properly extracted at each generation. Results of the application of our approach to the well-known Mackey-Glass chaotic time series dataset are shown and discussed.

PDF full paper