Explainable Machine Learning for Predicting Vehicle CO₂ Emissions: A SHAP-based Analysis

XXIV. IBANESS Congress Series on Economics, Business and Management, Ohrid, Makedonya, 11 - 12 Ekim 2025, ss.610, (Özet Bildiri)

Abstract: Carbon dioxide (CO2) emissions from road transportation represent one of the most significant contributors to

climate change and environmental degradation. According to the International Energy Agency (IEA), road transport alone

accounts for nearly one-quarter of global energy-related CO₂ emissions, making it one of the most carbon-intensive

industries. With the increasing number of passenger vehicles worldwide, emissions from cars have become a key driver of

environmental degradation, air pollution, and public health issues. Accurate prediction of vehicle emissions is critical for

developing sustainable policies and designing eco-friendly transportation strategies. In this study, we present an

explainable machine learning framework for predicting CO2 emissions from cars using a publicly available Canadian vehicle

dataset. Several regression models, including random forest (RF), gradient boosting machines (GBM), and support vector

regression (SVR), were evaluated to ensure robust predictive performance. To enhance model transparency, we

incorporated Shapley Additive exPlanations (SHAP), a state-of-the-art explainable artificial intelligence technique, to

interpret the contribution of each vehicle feature to the emission levels. The experimental results demonstrate that

ensemble-based models outperform traditional regression methods, achieving higher accuracy across MSE, RMSE, and R 2

metrics. This explainable framework not only enhances model fairness and accountability but also provides valuable

insights for policymakers, manufacturers, and consumers in achieving low-carbon transportation.

Keywords: CO2 emissions, explainable AI, SHAP, machine learning, sustainability.