A HYBRID ECONOMETRIC-NEURAL NETWORK MODELING APPROACH FOR SALES FORECASTING

Business sales forecasting is an example of management decision making in an ill-structured, uncertain problem domain. Due to the dynamic co mplexities of both internal and external corporate environments, many firms resort to qualitative forecasting techniques. However, these qua litative techniques lack the structure and extrapolation capability of quantitative forecasting models, and forecasting inaccuracies typical ly lead to dramatic disturbances in production planning. This paper pr esents the development of a hybrid econometric-neural network model fo r forecasting total monthly sales. This model attempts to integrate th e structural characteristics of econometric models with the non-linear pattern recognition features of neural networks to create a ''hybrid' ' modeling approach. A three-stage model is created that attempts to s equentially ''filter'' forecasts where the output from one stage becom es part of the input to the next stage. The forecasts from each of the individual sub-models are then ''averaged'' to compute the hybrid for ecast. Model development is discussed in the content of an actual sale s forecasting problem from a Danish company that produces consumer goo ds. Actual model performance is reported for a six-month time period. Knowledge gained from the modeling approach is placed in the context o f organizational learning about the nature of sales forecasting for th is particular company.