A New Parameter Determining Mechanism for Radial Basis Neural Networks

Neural network is a widely used and an effective artificial intelligence technique used for predictions and classifications which has been developed based on human biological neural system. Determining the structure of a neural network is a very complex task and there is no defined approach to determine the structure, especially the number of hidden nodes. Traditionally the number of hidden nodes is determined by carrying out a trial and error approach and by choosing the model which gives the best results. This approach takes a significant time and effort to determine the structure of a neural network. In this paper we propose a method to determine the number of hidden nodes using the cluster validity mechanism and to determine the parameters such as center and width of a radial basis neural network (RBNN) using the fuzzy clustering algorithm. Results shows that the proposed approach gives better classification rates comparedto the traditional approach for the data sets we used. Also it gives similar classification rates when compared with the best results reported in the literature. Therefore, the proposed mechanism provides an alternative to determine the number of hidden layers and other learning parameters during the training process.