The Introduction Of Combination Forecast Method And Empirical Analysis

In the process of actual forecasting, when the object to forecast is a complex system, only some effective information are provide form some way and ignore the other information, if we only choose the traditional individual forecast method. So J.M.Bates and C.W.J.Granger refer to a more scientific method, combination forecasting, which adopt different forecasting methods to predict the same problem and assemble them appropriately, then we may utilize fully useful information provided by different methods in order to improve forecasting precision.This paper studies on the principle and method of combination forecasting, and its application. Firstly, outline the basic theories related to forecast and studies on the predictive definition and the predictive steps. Secondly, this paper makes an all-sided for studies of several commonly-use traditional individual forecast methods, such as, multi-variant regression method, time serial method, gray system method and artificial neutral network method, as well as analyze and compare the advantages and disadvantages of all these method, bring out the idea of combination forecasting, and further discuss its method and weight of every method. Then, by the forecast object, this paper selects three individual forecast methods to predict which are liner regression, index smoothness and GM(1,1) forecast method, combines these three individual forecast methods together to build the final combination forecasting model under the criteria of minimizing the error sum squares, which is given in this paper, and apply this kind of model to forecast. Finally, analyze the results, which shows that we can take the model as an effective tool to predict about some complex systems.