An Automatic System for Sale Forcasting with Memory Reduction Technique

Accurate forecast of sales can help retail enterprises to improve production efficiency and competitiveness. In this paper, the sales forecast of Wal-Mart is realized based on LightGBM framework, which is an optimized GBDT model combining GOSS and EFB technologies. Considering the large and chaotic amount of data, we firstly compressed the historical data in memory, and then preprocessed the data with features, removing some features irrelevant to the model, and then extracted and classified the features, so as to obtain some statistics that can reflect the prediction accuracy of the model. Experiments show that the RMSE of Light GBM model is 2.07, which is better than that of the traditional linear regression model and SVM model.