In present, the prediction based on time series is applied in many fields, for example, weather forecast, stock, economics, communications and so on. As thorough study of mathematical theories and the development of application mathematics, people found that some of the practical data under dealing could be scored by random time series, but some of them had to be modeled to chaotic time series. So we should not only use statistical methods to forecast time series, but should estimate the time series whether random or chaotic, then choose the best method. The paper introduced several models of random time series, and the predicting methods based on these models. This paper also studied the prediction based on chaotic time series, and used the add-weighted local-region one-step model to forecast the Chengdu living price index, and compared it with the Exponential Smoothing, the result indicated this forecasting method was better than the Exponential Smoothing.
|