In this thesis, several types of time series models and their qualities and characters were introduced first. Among these, the stationarity and the invertibility were specified. The process of simulating dynamic data were summarized, which could be fell into six steps, they're the pre-dealing process of dynamic data, the choosing and recognizing of the model's form, the initial estimating of the model's parameter, the subtle estimating of the parameter, the judging of the result and the testing and mending of the model. Also two kinds of prediction methods--stationary linear least square prediction and new information prediction were compared, and the gradual consistency of them was proved. Finally the theory was used into the real world. A ARIMA model was established on the groundwater static level of Xijiao well fields of Anshan. And it was proved that the imitation was good.
|