Automatic Fuzzy Rule Extraction Based on Fuzzy Neural Network

摘要

In this paper, a hybrid algorithm based on tabu search (TS) algorithm and least squares (LS) algorithm, is proposed to generate an appropriate fuzzy rule set automatically by structure and parameters optimization of fuzzy neural network. TS is used to tune the structure and membership functions simultaneously, after which LS is used for the consequent parameters of the fuzzy rules. A simulation for a nonlinear function approximation is presented and the experimental results show that the proposed algorithm can generate fewer rules with a lower average percentage error.