Abstract:

This paper proposes to use a new type of radial basis function (RBF) network to approximate the functional coefficients of the state-dependent autoregressive model. The output weights of the new type RBF network, instead of constant parameters normally used, are the linear regression functions of the input variables. A fast-converging estimation method is applied to optimize the parameters of the model. Two benchmark chaotic time series, Mackey-Glass time series and Lorenz attractor time series, are used to test the performance of the proposed model. Simulation tests show that the predictive accuracy of the model is much better than that of other existing models.