Fuzzy neural networks with application to sales forecasting

Sales forecasting plays a very prominent role in business strategy. Numerou s investigations addressing this problem have generally employed statistica l methods, such as regression or autoregressive and moving average (ARMA). Sales forecasting is very complicated owing to influence by internal and ex ternal environments. Artificial neural networks (ANNs) have also been recen tly applied to learn the time series data since their promising performance s in the areas of control and pattern recognition. However, further improve ment is still necessary since unique circumstances, e.g, promotion, cause a sudden change in the sales pattern. Thus, this study utilizes fuzzy logic which is capable of learning (fuzzy neural network, FNN) for in order to gr asp the experts' knowledge. The proposed forecasting system consists of fou r parts: (1) data collection, (2) general pattern model (ANN), (3) unique p attern model (FNN), and (4) decision integration (ANN). Model evaluation re sults indicate that the proposed system can more accurately perform, than t he conventional statistical method and single ANN. (C) 1999 Elsevier Scienc e B.V. All rights reserved.