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

The prediction of chaotic time series has very abundant and profound meaning and can be widely applied. With a study on the relevant theory and methods, this paper applies BP neural network in the prediction of chaotic time series and its validity and universality are proved by matlab program.The paper can be summed up into as the following parts:1)Chaotic time series is introduced and its predictability, methods and scales are mainly discussed as well.2)Phase space reconstruction is an effective method to deep information mining of time series. In this paper, time delay and embedding dimension for chaotic time series analysis in phase space reconstruction as well as the estimated Lyapunov are thoroughly discussed and the algorithms is given accordingly.3)Neural network theory is applied to the prediction of chaotic time series while the realization process of the function by the BP neural network to be reconstructed is shown. And by using the trained network in prediction, the method is proved to be valid by simulation and the specific example of application is given.4)Parameter enactment method of BP network is brought out while the method to optimize parameter is discussed as well.