Prediction of chaotic time series based on incremental method for Bayesian network learning

The prediction of Chaotic time series constitutes a hot research topic of chaos theory, it is widely used in signal processing and automatic control field. In order to sufficiently model time series, on the base of the theory of phase-space reconfiguration, using the advantages of incremental method for Bayesian network learning in dealing the uncertainty to build a nonlinear prediction model for the prediction of chaotic time series. The method is applied to a chaotic time series produced by Henon equation, and the experimental results show that our prediction models has better predictability and stability than K2 algorithm and SVD predictive models.