Abstract:
A construction algorithm of RBF networks based on changing width factor is proposed to improve classification accuracy and shorten convergence time of RBF network. On the basis of subtractive center and samples as widith factor
σ,so
σ can be updated self-adaptively with the optimization of clustering center. RBF model clustering algorithm of multi-support vector machine, changeless width factor of RBF network model based on Gauseian function and this method model are used to classify three large data sets named Breast Cancer, Wine and Vowel, and make comparison from classfication accuracy and convergence time. The results show that this algorithm can greatly improve classfication accuracy and convergence speed.