Study on prediction modeling of the artificial neural network from the combination of multivariate analysis and mean generation function

A mixed prediction model involving a significant period of a predictand the physical factors influencing the change of predictand is built based on an artificial neural network from the combination of multivariate analysis and mean generation function. The new model is found to have a higher predicative accuracy (due mainly to the reasonable analysis for prediction method), when compared to results from multivariate statistical model or mean generating function prediction model.