The Prediction Of Chaotic With Noise Based On Neural Network

Time series prediction became more and more popurler in artificial intelligence and data mining researching. Using the method of nonlinear in time series prediction can solve practical engineering application' s problem. This paper mainly studied the characteristics of chaotic time series, as well as in practical applications of denoising and prediction.In current researching, the method of denoising in chaotic time series is linear method, however it loss the property;the nonlinear methed is better but complicated. On other hand, using the equation to predict time series not as accurate as neural network , use which different model will cause different prediction results.To slove these problems, this paper based on equation and actual sequence experiment, using the method and integration neural network model improve the denoising effect and prediction accuracy.In this paper, the author used sequences were generated by equations and application century, did a lot of experiments. The results include the following two parts: for the selection of neighborhood, this method firstly uses Hilbert-Huang transform to estimate the initial neighbour radius and the threshold limit the neighbourhood adaptively in the phase space; What is more, non-orthogonal projective approach is used to different neighborhoods. In the prediction model, both SOFM network and RBF network are used to improve the forecast accuracy. The numerical experiments results show that this method can better correct the position .