Stacked Ensemble Learning for Predicting Carbon Emissions Throughout the Life Cycle of Buildings

Aggressive urbanization and the imperative to achieve sustainable infrastructure bring attention to carbon-emission anticipation and reduction across a building's lifespan. The use of machine learning (ML) methodologies for robust emissions estimation at phases like material manufacturing, construction, usage, upkeep, and destruction is considered here. These are typical methods with traditional models depending on computation or single algorithm use, yet lacking in elasticity and encountering trouble in nonlinear dependencies as well as in high-dimensioned information. To overcome these challenges, a stacked ensemble model that integrates Random Forest (RF), eXtreme Gradient Boosting (XGBoost), and Support Vector Regression (SVR) is suggested. This ensemble method utilizes the advantages of both tree-based and kernel-based algorithms to identify global patterns and nuanced nonlinear trends. The ensemble is trained on actual building datasets and outperforms traditional models such as Linear Regression, Decision Tree, and simple ANN in accuracy and stability. It is 96.3% accurate and has a 94.8% F1-score, and it shows scalability across different building types and lifecycle phases. This data-driven, ML-based method enables data-informed, low-carbon decision-making during design and planning, leading to environmental sustainability in construction.