Research and application of chaotic time series prediction based on Empirical Mode Decomposition

Time series that composed of disperse observation like climatic time series have nonlinear and nonstationary features. Because of the superiority of Support Vector Machine in solving nonlinear problem and the advantage of Empirical Mode Decomposition in handling nonstationary signal, this paper combined the two methods in the research on chaotic time series prediction, and applied it to the seasonal precipitation forecast in Guangxi Zhuang Autonomous Region. Apart from this, this paper compares this result with RBF neural network algorithm and Support Vector Machine algorithm neither with the Empirical Mode Decomposition algorithm. Results show that relative to the directly predict methods, algorithm in this paper has the higher precision in prediction and better generalization ability.