Study On A Combination Forecasting Method Based On GEP

Grange and Bates proposed combination forecasting theory which aimed completely using the information mined by the single model forecasting theory. During the past several decades, combination forecasting theory is already gradually maturing. Many scholars'researches show that combination forecasting is more reliable and accurate than single model forecasting theory, so forecasting methods research has important theoretical and practical value.Linear combination Forecasting is a common mode in combination Forecasting. But it is relatively simple, and can't accurately show the complex systems'features, that will bring bad affect to the forecast accuracy. However, the nonlinear combination forecasting method can accurately show the complex systems'features. The key of nonlinear combination forecasting is to find the nonlinear combination forecasting function, but it is very hard to find the nonlinear combination forecasting function in traditional math method, this situation limits the application of the nonlinear combination forecasting method. So we introduce GEP to solve the difficult problem of finding the nonlinear combination forecasting function.GEP is a new concept which is developed on the base of the Genetic Algorithm and Genetic Programming. It overcomes the lack of GA and GP, and is more suitable for mining the relationship between the functions. For this reason, we will introduce GEP to the combination prediction, using its strong ability to function mining and the combination of nonlinear predictive function, to do our combination forecasting experiment.This article includes the following contents:●Analyzed the principle of the prediction, summarized some methods to evaluate the prediction results, and briefly introduced some single model used in the experiments;●Introduced the basic principles of forecasting and the existing forecasting approaches, focusing on comparing the linear and nonlinear combined forecasting;●Summarized the relationship between GA,GP and GEP, and compared them. Introduced the characteristics of GEP and the principles of excavation, analyzed its feasibility to mine of the nonlinear function in theory. On these basis, proposed the theory, methods and steps of combination prediction by use of GEP. ●Used CPI for forecasting experiment. The result shows that the Combination Forcasting of GEP can improve the prediction accuracy. The last part of this article analyzed and summarized the GEP prediction method.