The Research On Single Forecast Model Selection Oriented By Combination Forecast

With the development of economy, statistical prediction technology is also in the development of mature gradually. At present, more and more phenomenon is manifested in the form of the specific data. Therefore, statistical prediction technology has been applied in many fields. From the agriculture to the industry, from the population to the economy, from the energy to the development of society, from the transportation to the medical and health care, statistical prediction technology has been applied.According to the same social economic forecasting problem, because different decision makers take different point,they can choose different prediction methods.But sometimes,single forecasting model have been unable to meet our requirement for accuracy,so combination forecast has been used more and more.However,when choosing single forecasting model for combination forecast,there are many problems need to be solved.This article is about how to choose single forecasting model for combination forecast model.At first, the paper espounded the research background and research significance, reviews the development of the research literature,and explain the main contents,the research framework and research methods.Second,discuss something about statistical prediction technology,clear the detailed steps of forecasting work,outline the related theory of single forecast model,combined forecasting model,and the prediction effect evaluation index system.Then,analysis the choose of single forecasting model for combination forecast from the angles of qualitative and quantitative.Qualitative angle includes the choice of principle,Prediction purpose,Data changes, model applicability.Quantitative point of view, including the determination of single forecasting model number,the weight.And selected China’s private car ownership data from 1994 to 2013,make models,validate the rationality and feasibility of single forecasting model selection steps. Finally,analyse predicted results form predictive value and error value.lastly,gives Suggestions for the use of making combination forecast.