Application And Research On Combined Forecasting Model Based On Grey System And Neural Network Theory

Grey combined forecasting model is one of the main research directions in the combined forecasting theory, and the organic integration of the grey model with other models is an important content in the grey combined forecasting model. In this paper, the basic concepts of the grey prediction model and artificial neural network are presented, based on their modeling mechanism, the combination of grey GM (1,1) model and BP neural network as well as the grey Verhulst model and BP neural network are studied, three combined forecasting models are established for accuracy enhancement of prediction:(1) The combined forecasting model GM-BP1based on grey GM(1,1) model and BP neural network theory:Firstly, the error sequence is obtained by GM(1,1) model using original data sequence, and then in order to gain a forecasting error sequence, the BP neural network is built up to train the regression of error sequence. Finally, a new forecasting value is obtained while summing the GM(1,1) model prediction and forecasting error sequence. This new model correct the error of GM(1,1) model prediction using BP neural network, and its accuracy has been significantly improved.(2) The combined forecasting model GM-BP2based on grey GM(1,1) model and BP neural network theory:This model uses the partial-data sequence of the original data to a create partia-data GM(1,1) model group, and build a BP neural network to establish the nonlinear relationship between the fitted values and original data, the generated network estimates the forecasting development trend of the partia-data GM(1,1) model group, and achieves better results in the medium-and long-term forecast.(3) The combined forecasting model Verhulst-BP based on grey Verhulst model and BP neural network theory:The combination mode of grey GM(1,1) model and BP neural network, in which appeared the GM-BP2model is extended to the combination of grey Verhulst model and BP neural network. This model is not only able to handle a S-type or a single-peak-type sequence, but also has better prediction accuracy than the original Verhulst model.The application demonstrates that the three models are valid and acceptable.