Extracting compact T-S fuzzy models using subtractive clustering and particle swarm optimization

This paper presents a two-stage approach to extract compact Takagi-Sugeno (TS) fuzzy models using subtractive clustering and particle swarm optimization (PSO) from numeric data. On the first stage, the subtractive clustering is employed to partition the input space and extract a fuzzy rules base. On the second stage, the PSO algorithm is used to search the optimal membership functions (MFs), consequent parameters and the rule weights of the crude model obtained on the first stage simultaneously. Simulation results on two benchmark modeling problems show that the proposed approach is effective in finding compact and accurate TS fuzzy models.