Research On Combination Forecasting Method And Application In Forecasting Of Agriculture Machinery Total Power

Recently, with the development of society and economy and the requirement of management, more accuracier prediction is also demanded. So,the aim of this paper committed to improve the accuracy of combination forecasting and the important research content of this paper aimed in perfecting the theory of combination forecasting,then,taking those investigative theories into forecasing ang to guide practice activities. The important questions of previous studies were the weight of the single forecasing methods in combination forecasting,but,the choice of single forecasing methods mainly relyed on existed experience or obvious characteristics that particular process presented. So,in order to explore methods of improving combination forecasting accuracy, this paper would start from two aspects of the choice of single prediction methods and weight calculation methods of combination forecasting model under different principles,and taked the research theories into forecasing agriculture machinery total power in Heilongjiang province.The main content of the paper was as follows:Firstly,after searching related literatures,defined the status quo at home and abroad of combination forecasting.summarized some existed single prediction methods and analyzed the exsited problems and effect factors of combination forecasting.Secondly,systemically analysed the selection principles of single prediction method,choice principles and three screening procedures of single prediction method through the study were offered:ADF unit root test, selection for cointegration and removing redundant methods.Then, Using the software of Eviews5.0to analyze the specific example by way of ADF unit root test、 selection for cointegration and removing redundant methods. By contrast,it verified that the screening theory is effective,at the same time, also proved the necessity and importance of choosing single forecast models.Finally, from the set idea of mathe,made the useful information as quantify information and calculated the unique information amount of single forecasting models.Then,respectively established the traditional combination forecasting model and the IOWA operator combination forecasting model based on the unique information amount under three principles of square error sum、absolute error sum、curve difference. By evaluating and comparing for two types models in the evaluation index of combination forecasting model accuracy, it proved that IOWA operator combination forecasting model based on the unique information amount had more fitted accuracy.Putted investigative selection theory into predicting agriculture machinery total power in Heilongjiang,two types models which were composed of single forecasing methods after rigorous choice under three principles were obtained. The result showed that IOWA operator combination forecasting model based on the unique information amount could effective improve the fitted accuracy and prediction result,prediction value was more credible.This paper will offer systematic theory direction for the choice of single forecasting models, and offer the reference for improving the accuracy of combination forecasting.