Research on cross-border e-commerce customer churn prediction based on enhanced XGBoost algorithm with temporal-spatial features

Abstract

This research addresses the critical challenge of customer churn prediction in cross-border e-commerce by proposing an enhanced XGBoost-based framework that integrates temporal-spatial features and dynamic weight adjustment mechanisms. In response to the complex characteristics of international e-commerce, including regional behavioral variations, seasonal patterns, and logistics impacts, this study develops novel approaches to feature engineering and algorithm optimization. The enhanced model incorporates continuous temporal processing, adaptive weight adjustment, and business rule-based feature interactions to achieve superior prediction performance. Through extensive experimentation with large-scale cross-border transaction datasets, the research demonstrates significant improvements in prediction accuracy across diverse geographical regions while maintaining model interpretability. The findings contribute substantially to both the theoretical advancement of machine learning applications in cross-border e-commerce and practical implementations in customer relationship management. The proposed framework provides valuable insights for international e-commerce platforms seeking to implement more effective customer retention strategies and optimize their operational efficiency in global markets.