A hybrid method for identifying T-S fuzzy models

This paper presents a hybrid approach to extract compact Takagi-Sugeno fuzzy models from numeric data, using subtractive clustering (SC), particle swarm optimization (PSO) and least square method. The feature of this method lies in the following: (1) the input space is partitioned and initial fuzzy rule bases are extracted by SC; (2) the optimal parameters of the membership functions are evolved by PSO, based on the initial fuzzy rule bases; (3) the consequent parameters of the rule base are analytically derived by the Moore-Penrose pseudo inverse, instead of being iteratively tuned. Simulation results in Mackey-Glass chaotic time series prediction and nonlinear plant modeling problems show that the proposed method is effective and robust in finding compact and accurate T-S fuzzy models.