A neural-network-based forecasting algorithm for retail industry

To obtain the inherent laws from large amounts of data records in retail industry and to provide valuable information for retailers, this paper presents a neural-network-based forecasting algorithm, which adopts Holt-Winters´ model and a neural network. Different from traditional forecasting algorithms, this algorithm rearranges Holt-Winters model, and builds a neural network on it. Furthermore, it puts forward a training algorithm to optimize the adjustable neural network weights by minimizing a defined cost function, which has greatly improved the forecasting accuracy. Experimental results at the end of this paper also prove the superiorities.