Abstract: This paper, aiming at the nonlinear characteristics of CPI and the parameters of being difficult to be objectively determined in RBF neural network, puts forward a kind of optimization of RBF neural network method which combines orthogonal least squares (OLS), K-means clustering and gradient descent algorithm, computing activation function with the linear combinations of Gauss, reflected sigmoidal and inverse multiquadrics radial basis functions, then builds a model for CPI to fit and forecast by using the optimization algorithm of RBF neural network. Experimental results show that the model is of a good convergence and generalization ability, the model has a certain universal applicability in the prediction performance which is obviously superior to the single method forecast and the hybrid network on Gauss kernel function.


Key words: RBF neural network, optimized hybrid algorithm, CPI forecasting