ML-ERV: A machine learning-based CO₂ emissions model designed for rental vehicles

Abstract On-road carbon dioxide (CO 2 ) emissions from rental vehicles are very much affected by vehicle category, types of vehicle transmission, fuel type and traveled distance. In this paper, four machine learning models (decision trees, random forest, AdaBoost and XGBoost) were used to predict CO 2 emissions from rental vehicles. The models were trained and tested on Carbookr’s bookings dataset. The results show that the traveled distance, the included mileage, the ACRISS code, the rental duration, the driver’s home country, the country where the driver works and the country where he pickups up the vehicle as well as the country where he returns the vehicle had clear impacts on the transient CO 2 emission rates. There was a positive correlation between these features and CO 2 emission rates. The findings also indicate that XGBoost achieves the best accuracy performance when compared to other machine learning algorithms in terms of the Root Mean Squared Error (RMSE), the Mean Absolute Error (MAE), the Mean Absolute Percentage Error (MAPE) and the Coefficient of Determination ( R 2 ). The results show that the machine learning-based CO 2 emission model ML-ERV trained using XGBoost algorithm outperforms the state-of-the-art CO 2 emissions calculators.