RBF-Based Divide-and-Cooperate Time Series Forecasting Model And Its Computer System

Time series analysis is an analyzing approach for stochastic data series analysis and constructing its mathematical model and has extensive application in the field of economy and finance. At present the analyzing approaches for linear time series model such as the models of auto-regressive, moving average and autoregressive-moving average are successful and the computer software has been presented. But there is no development in nonlinear model of time series. Especially, the existing approaches are of low efficiency for time series with high order.The application of artificial neural networks to time series model especial to nonlinear model has become research highlight. Due to its simple structure, quick converging speed and excellent approximation ability, recently, RBF has many successful applications in the model of time series. This paper constructs a Divide-and-Cooperate model based RBF for forecasting time series with its computer system. The major work is listed as follows:1. Presenting a method to decide the order of time series via RBF. This method is constructed under the criterion of learning mean square error.2. Constructing divide-and-cooperate forecasting model based RBF for time series. In this model the input of a variable-separable time series is decomposed. The divide-and-cooperate forecasting model is of excellent approximation ability in the time series with high order.3. Designing the computer system of divide-and-cooperate forecasting model based RBF for time series.