Superstore Sales Forecasting Based on Elastic net Regression and BP Neural Networks

Abstract

Accurate sales forecasting is an important guide to business operations. It allows the operations back office to allocate resources to assist managers in making decisions. However, from the existing sales data of the store, it summarizes the change law of commodity sales, and dynamically predicts the sales in the future for a period of time according to the law. Moreover, this paper uses the elastic regression network model and BP neural network model to predict the sales of shops over a period of time. In order to improve the accuracy of the model, the model data is combined with one-hot coding. MAP, MPE and RMSE were chosen to be used as computational metrics for the evaluation for quantifying the accuracy of the mode. A comparison of the performance of the two models is made, which in turn has practical implications for companies to improve their promotions and increase their revenue.