Sales Volume Forecast of Sports Fashion Clothing Based on Machine Learning Algorithm A LGBM-SHAP Approach

The forecast of sales volume is of great significance to the development of e-commerce. In this paper, a sales forecasting model based on LGBT and SHAPE algorithm is proposed. The LGBT model is used to train the forecasting model, and the SHAPE model is used to quantitatively analyze the contribution of each feature, which not only retains the good forecasting effect of machine learning, but also improves the interpretability of the results. Based on the daily frequency data of a sports brand, this paper finds that the prediction effect of the model is very good, and the model also finds out that the transaction conversion rate, the number of visiting users, the page views and the average price of a single product are the main factors affecting sales, which provides some reference for improving the explanatory power of the machine learning model in finance.