The Research Of Chaotic Time Series Prediction Method Based On The BP Neural Network

Chaos Theory had researched for a long time in the history, and the forecast of chaotic time series is an important branch of the Chaos Theory. The methods of the forecast have been well developed. Today there has been a hot issue that Neural Networks has been used to the prediction of chaotic time series, as a new information processing science.It is one important means to analyze and observe the evolution of time series. Since 1960s people found that time series have chaotic characteristics, which are from astronomy, hydrology, meteorology and other fields, such as Sunspots, Runoff, Precipitation, etc. Nonlinear Analysis Method for Dynamic Modeling has been proposed in the Chaos Theory, Such as phase space reconstruction, in order to study and deal with chaotic time series. Chaotic prediction is that these specific applications of nonlinear analysis. It is better than Statistical forecasting methods of stochastic systems and linear regression linear regression. In recent years, Nonlinear modeling tools on neural networks, basing on the theory about phase space reconstruction of chaotic systems support vector machines; the application has created the conditions for discovering and studying the internal laws of chaotic systems.This paper studied the prediction method of chaotic time series which Combined with the BP neural network theory and methods. The research included the following:For simple chaotic systems, It analyzed and confirmed that the character of chaos. The advantages and disadvantages of the traditional prediction of chaotic time series and analysis are discussed; In regard to the shortcomings of the BP neural network. It proposed the improved method about Momentum factor, Learning rate and Gradient Descent Algorithms to enhance the accuracy of neural network prediction. Before the forecast, the Chaos data are denoised and standardized. This paper developed the prediction system of Improved BP neural network Which applied to forecast the sunspot chaotic time series Practice has proved that, the prediction system of Improved BP neural network has the better prediction accuracy than the classical neural networks method, chaotic global prediction method and the local prediction method.