A Nonlinear Combination Forecasting Method Based On Fuzzy Logical System And Its Application

First presented by J. M. Bates and C. W. J.Granger, the research ofcombination fOrecasting method has got great development. EarIy combinationof forecasts is based on the Iinear combination of the resuIts of several singlemodels. However, this method isn't usabIe when there are sorne complicatedrelationships betWeen the real value and the prediCted results of single modeIs.Under the review of the traditionaI forecasting models, a combinationforecasting method based on fuzzy logicaI system is presented in this paper.The main idea of this method is that the combination funCtion gh(X) can besimulated by Takgi-Sugeno fuzzy model which is determined by the backpropagation Iearning aIgorithm fOr the universaI approach property of the fuzzysystem. Not only the difficuIty of construCting &(X) is reduced but aIso thepredicted precise is improved by the new combination forecasting method.FinaIIy, the method is appIied in the research of Jiangsu telecommunicationdemand prediction during 2001 -- 2005.A nonlinear model is construCtedthrough the learning and verification of the simuIation results of the nearestdecade teIecommunication information from singIe models. The comparison ofthe resuIts of this modeI, traditional optimal combination model and everysingIe model shows that the resuIt of this nonlinear combination method isobviously better than traditionaI fOrecasting models. Both theoreticaI analysisand the forecasting example t6stify that the new combination forecastingtechnique is feasibIe and effective.