A combination forecasting model to chaotic time series

Chaotic time series exist in many natural economic phenomena. The commonly used forecast methods including adding-weight one-rank local-region method and forecast method based on the maximum Lyapunov exponent. The first method which could cause the forecast showing a smooth trend and the forecast result of the second method may have a drastic change in the trend. So the scope of application of these two methods is different. In the prediction of road day traffic time series, the time series is smooth overall and it also contains rich volatility. For this characteristic, a combination forecasting model is proposed in this paper based on organic combination of the two methods. It can solve the determination of the embedding dimension in chaos forecast and the evidence showed that the prediction is effectively.