Machine Learning and Big Data Frameworks for Sustainable Economic and Environmental Decision-Making

Abstract

Climate change and environmental pollution have led to the increased need for developing sustainable economic and environmental models that could be capable of considering the intricate relationships between economics, energy, industries, and carbon dioxide emissions. This study developed and evaluated machine learning and big-data frameworks for greenhouse gas emission prediction and sustainability-oriented decision support. A quantitative predictive approach was applied using a multi-regional and multi-sector dataset containing 179,185 observations from 1995 to 2022. Greenhouse gas emissions were modeled using Value Added, Employment, Energy Carrier Net Total, Region, Sector, and Year. Linear Regression, Random Forest, and XGBoost were implemented and evaluated using Mean Absolute Error, Root Mean Square Error, and the coefficient of determination (R²). Feature importance and SHAP analysis were additionally performed to improve model interpretability and transparency. The results demonstrated that XGBoost achieved the highest predictive performance, with MAE = 0.6007, RMSE = 0.9172, and R² = 0.9179, outperforming both Random Forest and Linear Regression. Energy consumption emerged as the strongest determinant of greenhouse gas emissions, while electricity production, waste management, petroleum refining, and livestock-related sectors significantly contributed to emission variation. The results show that the machine learning ensembles deliver accurate and understandable systems that can be used for environmental-economics prediction and sustainable decision making. This proposed system may aid in carbon-based planning, sectoral sustainability evaluation, and environmental policy formation.