Construction of a health food demand prediction model using a back propagation neural network

Abstract

For business operations, determining market demands is necessary for enterprises in establishing appropriate purchase, production and sales plans. However, many enterprises lack this ability, causing them to make risky purchasing decisions. This study combines a back propagation neural network and the Particle Swarm Optimization Algorithm (PSOBPN) to construct a demand prediction model. Using a grey relational analysis, we selected factors that have a high correlation to market demands. These factors were employed to train the prediction model and were used as input factors to predict market demands. The results obtained from the prediction model were compared with those of the experiential estimation model used by health food companies. The comparison showed that the accuracy of PSOBPN predictions was superior to that of the experiential estimation method. Therefore, the prediction model proposed in this study provides reliable and highly efficient analysis data for decision-makers in enterprises. © Maxwell Scientific Organization, 2013.