Hierarchical genetic fuzzy systems

This paper introduces a hierarchical evolutionary approach to optimize the parameters of Takagi-Sugeno (TS) fuzzy systems. The approach includes a lea st-squares method to determine the parameters of nonlinear consequents. A p runing procedure is developed to avoid redundancy in each rule consequent a nd to achieve proper representation flexibility. The performance of the hie rarchical evolutionary approach is evaluated using function approximation a nd classification problems. They demonstrate that the evolutionary algorith m, working together with optimization and pruning procedures, provides stru cturally simple fuzzy systems whose performance seems to be better than the ones produced by alternative approaches. (C) 2001 Elsevier Science Inc. Al l rights reserved.