An Improved Multi-Objective Genetic Algorithm for Tuning Linguistic Fuzzy Systems

M. J. Gacto, R. Alcalá and F. Herrera

This work proposes the use of Multi- Objective Evolutionary Algorithms to obtain Fuzzy Rule-Based Systems with good accuracy-interpretability trade-off. To do this, we present a new post-processing method that performs rule selection and membership function tuning by focusing in the Pareto zone containing the most accurate solutions but with the least number of possible rules. This method is based on the well-known SPEA2 algorithm, applying an intelligent crossover operator, considering some modifications to concentrate the search in the desired Pareto zone and including an incest prevention mechanism in order to obtain more global optima. The results show that improving the trade-off between exploration and exploitation in the search process enhances the SPEA2 algorithm performance.

PDF full paper