he expansion of electric vehicles in Brazil demands robust charging infrastructure and efficient applications to support users. In this context, customer retention represents a strategic challenge, considering that mobile applications present high abandonment rates and that the cost of acquiring new customers exceeds that of retention. Churn prediction using Machine Learning techniques emerges as an approach to identify users with propensity to abandon, enabling retention interventions. This work compares the performance of XGBoost and Random Forest algorithms in churn prediction for electric vehicle charging applications, simultaneously evaluating the impact of different data balancing techniques. The results demonstrated better performance with XGBoost and revealed that balancing techniques did not provide significant improvements in robust algorithms.