Global prediction for chaotic time series based on continued fractions

A non-linear global predictable model for chaotic time series is built, and it based on continued-fraction approximation and phase-space reconfiguration, while the dynamic model can't be known for the system of chaotic time series, we develop the polynomial approximation method to the rational-fraction approximation's in theory, and we can analyze the characteristic and gain the prediction by substituting the model for the equation of the chaotic time series. The results of theoretic analysis and computer simulation have proved this method is practically feasible. We can predict it accurately comparatively and get an explicit expression.