A Comparative Analysis of Machine Learning Algorithms for Reliable Range Prediction in Electric Vehicles

Electric Vehicles (EVs) represent a rapidly expanding industry that has seen significant advancements in recent years. As the number of electric vehicles continues to increase, critical issues such as the Remaining Driving Range (RDR) reliability require attention. This research article explores the reliability, accuracy, and efficiency of various machine learning algorithms employed to predict the RDR of EVs. A comparative analysis is conducted among Multiple Linear Regression (MLR), Gradient Boosting Decision Trees (GBDT), XGBoost—a refined and efficient variant of GBDT— and the Markov Decision Process (MDP). These algorithms were evaluated using data collected under controlled laboratory conditions. The study concludes that XGBoost exhibits the most favorable training time and accuracy balance. In contrast, GBDT and MDP demonstrate slower performance due to the inherent complexity of their tree-based structures. MLR emerges as a strong contender, offering a close second in terms of accuracy and training speed. Overall, XGBoost is identified as the most effective algorithm for the given dataset, while both XGBoost and MLR benefit from more optimized data structures compared to GBDT and MDP, which rely on post- pruning techniques.