Comparative Analysis of XGBoost Method with Gradient Boosting for Vehicle Carbon Emission Prediction

Abstract

Predicting carbon dioxide (CO₂) emissions from vehicles is a crucial effort in mitigating the environmental impact of the transportation sector. This study compares the performance of two boosting models, XGBoost and Gradient Boosting, in predicting vehicle CO₂ emissions. The dataset includes various technical features of vehicles such as engine size, fuel consumption, and transmission type. Data preprocessing steps involved normalizing and encoding categorical features to ensure data readiness. XGBoost and Gradient Boosting models were implemented with an 80:20 data split for training and testing. Model performance was evaluated using metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R-squared (R²). The results indicate that Gradient Boosting slightly outperformed XGBoost with MAE of 2.092, MSE of 14.414, RMSE of 3.796, and R² of 0.9958, compared to XGBoost which achieved MAE of 2.098, MSE of 14.697, RMSE of 3.833, and R² of 0.9957. Both models demonstrated excellent performance, with Gradient Boosting being more accurate in predicting CO₂ emissions. These findings provide significant insights for the development of environmental policies and the design of more eco-friendly vehicles.