Research On Prediction Modeling Of Multivariate Series Chaotic Time

In recent years, the nonlinear system, especially chaotic time series analysis isgetting more and more attention. This paper use artificial intelligence method torealizing the reasoning and prediction for uncertain information.Firstly, we use the advantages of Bayesian networks (BNs) in dealing theuncertainty together with the theory of phase-space reconfiguration to build a nonlinearprediction model for the prediction of single variable chaotic time series. In this paper,we can analyze the dynamic character and realize the prediction.Then, with analysis of the underlying relationships among different state spaces,the connections among multivariate time series are discussed. The principal componentsanalysis (PCA) method is used to extract the joint information of multiple variables in acomplex system. On the basis of this, a new methodology is constructed to model andpredict multivariate time series.Finally, the prediction model is validated by simulation. The experiment resultsshow that our prediction model has good predictability and stability, and can predict thechaotic time series effectively.