Simultaneously Structural Learning and training of Neurofuzzy GMDH using GA

This article presents a new approach for Structural Learning of Neurofuzzy (NF-) GMDH networks, based on Genetic Algorithm (GA) optimization. Conventional methods, prune unnecessary links and units from the large network by minimizing the derivatives of the partial description. In proposed method pruning of needless links, units and fuzzy rules in each partial description, has been done by adding some extra binary weights to the conclusion part of each partial description. Two kinds of GA also proposed, necessary fuzzy rules in the conclusion part of each partial descriptions in NF-GMDH network, are chosen by using the binary-coded GA, and system parameters are adjusted by using the real-coded GA. Finally the newly proposed method is validated in classification of Iris data.