Forecasting Models Of The Multi-model Multi-method Research And Application

In the procedure of forecasting some complex systems, satisfactory results rarely appeared if we used only one forecasting method and one model. Therefore, based on the multi-level recursive method, a synthetic prediction pattern was developed. This pattern could reach a better forecasting result by composing advantages of several models and methods, and matching them. It is named with multi-method and multi-model synthetic prediction pattern.The influence of seasonal factor exists universally in the objective world. Seasonal factors have played great role in the history in agriculture, industry, commercial business, astrology, meteorology and other fields relative with human being. The time series which take month or season as observation unit is usually affected by seasonal factors and reveals periodical variation, which brings difficulties in researching and explaining the physical rule and significance of it. Thus, it is especially important to solve the problem of forecasting of seasonal time series.A profound theoretical analysis of multi-method and multi-model synthetic prediction pattern is achieved in this thesis paper, and based on that, a pattern of synthetic categorical prediction is developed in order to solve the problem in the paragraph above. In this thesis, the seasonal factor adjustment method X-11 is applied in the system of weather forecasting, which is usually used in economy. The X-11 method adjusts such data as monthly average temperature with seasonal factors, which helps to interpret the vary rule of it more precisely and to make it possible to get a precise model. The synthetic categorical prediction pattern in this thesis is the modeling and forecasting after the adjustment by use of X-11 method. It has been used in long-term weather forecasting system and gotten satisfactory results in forecasting the monthly average temperature and monthly precipitation of Xingkaihu Farm, Heilongjiang province. Besides, some model that consist of guide variable help to get better effectiveness in the forecasting of seasonal time series.