Sales Forecast of Retail Commodity on the Basis of LightGBM and Xgboost

With the development of market economy and economic globalization, enterprises encounter more and more market competition. On this basis, the product selection is crucial for enterprises, which strongly rely on sales prediction. In this paper, the LightGBM and Xgboost are adopted to predict monthly sales of retail products sold in a great number of shops after reasonably handling the abnormal missing data and setting the pre-processing of multiple data sets. By comparing and analyzing the prediction results, LightGBM model performs better than Xgboost model. At the same time, the key factors affecting the monthly sales are found including the number of shops, the types and quantity of items, the items reservation, the price of items, and the past sales. The prediction model and analysis results have a certain reference value for the actual sales prediction in companies’ operations. These results shed light on guiding further exploration of sales strategic arrangement and corporate strategy.