A novel rough neural network based on fuzzy partition

Abstract

A method to partition the universe of discourse based on fuzzy clustering is proposed to solve the partition problem in the process of constructing rough neural network. Considering traditional clustering algorithm has the problem of easily fall into local optimum, a modified PSO algorithm with crossover and mutation operators is combined with FCM algorithm. And a new fuzzy clustering algorithm (CMPSO-FCM) is proposed. The searching capability and clustering effectiveness are improved by this new algorithm. Then the fuzzy similar matrix, which is used for attribute reduction, is calculated by using fuzzy partition matrix and the definition of fuzzy similar measure after fuzzy clustering result is achieved. And a set of fuzzy rough decision rules are acquired by entropy method. Finally, a rough neural network is designed under these decision rules. Experiments results show that, compared with traditional rough neural network, this method has superiorities at the aspect of structure, classification precision and generalization.