A new prediction model was proposed by Newton interpolation and neural network based on time series. Firstly, history sales data were fitted by Newton interpolation and neural network was trained by mass data, which makes up for the deficiency of history data. Secondly, time series factors were fitted by substituting the trained neural network for traditional least-squares methods to obtain the predictor. Simulation result shows the fitting capacity and the prediction precision of the model is effectively improved.

Newton interpolation,    , Neural network,    , Time series,    , Forecast