An intelligent sales forecasting system through integration of artificial neural network and fuzzy neural network

Abstract

This paper attempts to develop an intelligent sales forecasting system which can consider the quantitative factors as well as the non-quantitative factors. The proposed forecasting system consists of four parts: (1) data collection, (2) general pattern model, (3) special pattern model, and (4) decision integration. In the general pattern model, a feedforward neural network with error-backpropagation (EBP) learning algorithm is employed to learn the time series data, or quantitative factors. However, unique circumstances, e.g., promotion, may cause a sudden change in the sales pattern. To this end, this paper utilizes fuzzy logic which is capable of learning (fuzzy neural network, FNN) to learn the experts' knowledge regarding the effect of promotion on the sales. Finally, the outputs from the above two mentioned models are integrated with time effect through a feedforward neural network with EBP learning algorithm. Evaluation of the model results indicates that the proposed system performs more accurately than the conventional statistical method and single artificial neural network (ANN).