ANN-ARMA Model for Forecasting Product Consumption Based on Non-Stationary Time Series

A new model of integrating artificial neural network (ANN) with auto regressive moving average (ARMA) is studied to handle existing problems of forecasting methods of product consumption based on non-stationary time series. Because the non-stationary time series can be divided into the certain and stochastic parts, the ANN-ARMA model is proposed. The certain part that is fitted by the ANN model denotes their non-stationary trend, and the stochastic part that is fitted by the ARMA model denotes their stationary and random component. The sum of forecast values of the ANN model and the ARMA model is considered as the optimal forecast value of future product consumption. A simulation example indicates the forecast precision of the ANN-ARMA model to be superior to that of the ANN model.