Rough Neural Network Modeling Through Supervised G-K Fuzzy Clustering

On the basis of fuzzy rough data model (FRDM), a method to construct rough neural network is proposed. By adaptive Gaustafason-Kessel (G-K) clustering algorithm, fuzzy partition can be accomplished in input data space. Then based on the search of cluster number, optimal FRDM will be found, and by integrating it with neural network technique, corresponding rough neural network is constructed. The experiment results indicate that rough neural network is superior to traditional Bayesian and learning vector quantization (LVQ) methods, moreover, rough neural network has more powerful synthetic decision-making ability than single FRDM model.