GNNFRM: Genetically constructed neuro new fuzzy reasoning model

Abstract

In this paper, a genetic algorithm with adaptive probabilities of crossover and mutation is introduced to find near global optimum parameters for the Neuro-new fuzzy reasoning model (NNFRM). The parameters to be optimized are those of input membership functions, output membership functions and relation matrix. A fuzzy evaluation criterion is introduced to evaluate the different fuzzy models. This criterion stresses the fact that the fuzzy system must be comprehensible and transparent to the user. The performance of the proposed model is evaluated using a benchmark problem. Also, the generalization of the proposed model is compared to the feed forward neural network. It is shown that the proposed GNNFRM outperforms other modeling methods. The generalization of the proposed model is better than that of the feed forward neural network.