A Research on Weight Acquisition of Weighted Fuzzy Production Rules Based on Genetic Algorithm

Abstract

In order to increase the knowledge representation power and improve generalization capability of fuzzy production rules (FPRs), a weighted FPRs (WFPRs), which incorporates the concepts of knowledge representation parameters (local weight, global weight and threshold), has been presented. However, the acquisition of these knowledge representation parameters is significant but difficult. This paper advances the weights acquisition of WFPRs by means of genetic algorithm (GA). First, a model based on GA is designed to obtain the local weights and the global weights of WFPRs; and then, the experiments on classification problem with continuous-valued attributes is performed. The experiments proved that the generalization capability of a fuzzy production system can be improved greatly by training and optimizing the weights with the use of GA, and verified the rationality and validity of the method sequentially