The forecasting method of selecting-best model based on artificial neural networks and its application

In order to deal with the problem of how to scientifically determine the weights of combined forecasting models(CFM),a new method is presented in this paper.According to the principle of the best selection,the combined forecasting model is transformed into a problem of pattern recognition,it can be resolved by the method of improved BP artificial neural network(ANN) which owns the ability of non-linear mapping.The given example shows that the so called selecting-best forecasting model not only successfully avoids the complex progress of computing the weights of combined forecasting models,but also owns the properties of clear concept,easy operation and good characters of the forecasting models.As a special case of variable-weighting CFM,it has some values in application.