The growing adoption of electric vehicles (EVs) underscores the need for optimizing charging infrastructure and cost management. Efficient charging strategies require accurate cost predictions to support users and policymakers in planning sustainable energy consumption. Machine learning (ML) techniques have shown promise in uncovering patterns within EV charging data, aiding in better pricing models and infrastructure management. However, determining the most suitable ML model remains a challenge due to varying data characteristics, feature dependencies, and performance disparities across algorithms. This research evaluates the predictive capabilities of several ML models, including Random Forest, Gradient Boosting, XGBoost, LightGBM, CatBoost, Support Vector Machines (SVMs), and K-Nearest Neighbors (KNN). A structured approach involving data preprocessing, feature engineering, and hyperparameter tuning is employed to enhance predictive accuracy. Furthermore, ensemble learning is explored to improve model generalization and reliability. The study reveals that ensemble-based methods yield the best performance, minimizing error while effectively capturing influential factors such as energy consumption, cost per kWh, and charging efficiency through SHAP analysis. The insights gained from this work are valuable for stakeholders such as energy providers, policy developers, and automotive manufacturers. Leveraging ML-driven insights can lead to improved charging infrastructure, more precise pricing mechanisms, and enhanced grid optimization. Future work could explore integrating real-time charging data and advanced learning techniques to refine cost prediction models further. This study contributes to the development of intelligent and data-driven EV charging strategies, promoting a more efficient and sustainable energy ecosystem.