Design of fuzzy classification system based on hybrid Co-evolution Algorithm

A novel approach to construct accurate and interpretable fuzzy classification system based on hybrid Co-evolution algorithm is proposed in this paper. The approach is composed of three phases: (1) the initial fuzzy system is identified using the Simba algorithm and the fuzzy clustering algorithm; (2) the fuzzy rule pool is optimized by the Michigan-style genetic algorithm; (3) the structure and parameters of the fuzzy system are optimized by the Pittsburgh-style Co-evolution algorithm. The hybrid Co-evolution algorithm has the advantages of Michigan-style and Pittsburgh-style algorithm. It owns three species including the number of fuzzy rules species, the premise structure species and the parameters species. Considering both precision and interpretability, the fitness function is calculated on cooperation of individuals from the three species. The proposed approach is applied to two benchmark problems, and the results show its validity.