A nonlinear time series method with neural networks
Wen Xinhui Chen Kaizhou
The nonlinear time series method is developing rapidly now, because many phenomena inthe real world can not be solved by the linear model.This paper first analyzes structure-mechanism about the time series model, and then aneural netwok is used to deal with the nonlinear signal process. We have structured a new non-linear time series neural network model.We have also compared the method with the AR model and the SETAR model. Using thethree methods , we have forceasted the number of sunspots and the results show that the neu-ral network method is better than the two others.