Research on User Churn Prediction of Traditional Supermarket Digital Platform

This paper uses the advantages of machine learning algorithms to conduct empirical research on classification prediction to explore the user churn prediction model of traditional supermarket digital platform. User reviews on the online platform of Wumart were used as the data source, and user information of Ehime Orange purchased was used as the research object. Starting from the two dimensions of user value characteristics and evaluation characteristics, the user's purchase information is collected, and five characteristic data such as the latest consumption time, consumption times, consumption amount, rating and review text are obtained by calculating the information. Later, a predictive model was built based on XGBoost, a machine learning algorithm, to predict the trend of user churn. By comparing and analyzing the contribution of characteristic variables in user churn prediction, the types of lost users are divided according to key characteristic variables such as average monthly consumption times and average consumption amount, and corresponding retention strategies are proposed. The validation found that the sentiment of the score and review text had a significant impact on user churn prediction. By analyzing the important variables affecting the prediction of user churn, this paper summarizes two types of churn users, and formulates corresponding retention strategies for users with less demand and no demand. This is of great significance for reducing the loss of digital platform users, maintaining old users, improving the profits of supermarkets, and achieving the goals of healthy and sustainable development of supermarkets.