A multi-model approach for predicting electric vehicle specifications and energy consumption using machine learning

The increasing reliance on electric vehicles (EVs) necessitates advanced predictive models to enhance performance and sustainability, especially against climate change driven by fossil fuel combustion. This study advances the field using cutting-edge machine learning techniques to model and predict various EV characteristics and performance metrics. Using a comprehensive dataset with parameters such as model year, make, model type, vehicle class, motor power, and energy consumption metrics, we tested classification and regression models to forecast categorical features (vehicle make, category) and continuous features (energy consumption, range, charge time, motor power). The models were assessed using key performance indicators like accuracy, recall, F1 score, precision, MAE, RMSE, and R². The random forest and Naive Bayes classifiers excelled in classification tasks, achieving accuracies of 95.58% and 100%, respectively. The Linear Regression model showed superior performance in predicting energy consumption in city driving conditions, with an R² value of 0.9982. The K neighbors regressor was most effective in predicting range and motor power, with R² values of 0.9800 and 0.9300, and the Huber regressor was most effective in predicting charge time. The study demonstrates the analytical models’ proficiency in projecting vehicular attributes and performance metrics by dividing the data into 70% for training and 30% for validation. The research provides crucial insights for policymakers, highlighting the potential of increased EV adoption to significantly decrease greenhouse gas emissions from the transportation sector, aiding in climate change mitigation.