Prediction of EV Charging Stations Congestion using XGBoost Approach

Abstract

This paper presents a data-driven approach for predicting congestion at electric vehicle charging stations using XGBoost. By combining real-world usage data with synthetically generated sessions and modeling personalized, time-based usage patterns, our method improves accuracy and interpretability. Cross-validation and residual analysis confirm strong performance comparing to similar models. This approach led to useful predictions that are practical and user-centered, and extends the possibilities for forecasting within emerging forms of mobility and energy systems.