A Multi-Model Approach to Carbon Footprint Prediction Using Machine Learning with Explainable AI

Estimating carbon footprints is essential for mitigating climate change and encouraging environmentally friendly practices, both at the level of individuals and society as a whole. This study proposes an extensive machine learning framework that predicts personal carbon footprints from lifestyle patterns such as transportation, energy usage, diet, and waste production. We benchmark the performance of five different model architectures Random Forest, XGBoost, Simple Neural Network, Deep Neural Network, and Ridge Regression against a range of different metrics. Ridge Regression achieves the highest accuracy of 98.7% with R² = 0.9832, outperforming complex ensemble and deep learning approaches on the test dataset. The system incorporates SHAP values for model interpretability, enabling users to understand feature contributions to their predictions. Feature importance analysis identifies plane journey count (11.54), personal vehicle distance (6.22), and public transport usage (3.79) as the most significant predictors, highlighting transportation's dominant role. The framework is implemented as a web application that offers explainable insights and realtime predictions. Our research shows that individual climate action can be effectively supported by machine learning and explainability, and that accuracy need not be sacrificed for model simplicity.