Explainable AI-Driven Analysis of Construction and Demolition Waste Credit Selection in LEED Projects

Selecting Construction and Demolition Waste (CDW) credits in LEED-certified projects is essential for sustainable building management, often requiring specialised expertise and contextual sensitivity. However, existing studies provide limited analytical insight into why certain CDW credits succeed or fail across different project contexts, and no explainable AI–based framework has been proposed to support transparent credit decisioning. This gap underscores the need for a data-driven, interpretable approach to CDW credit evaluation. This study proposes an explainable artificial intelligence (XAI)-based model to support CDW credit selection and to identify the key factors influencing credit performance. A dataset of 407 LEED green building projects was analysed using twelve machine learning (ML) algorithms, with the top models identified through Bayesian optimisation. To handle class imbalance, the SMOTE was utilised. Results showed that MRc2 and MRc4 credits had high predictive performance, while MRc1.1 and MRc6 credits exhibited relatively lower success rates. Due to data limitations, MRc1.2 and MRc3 were excluded from analysis. The CatBoost model achieved the highest performance across MRc1.1, MRc2, MRc4, and MRc6, with F1 scores of 0.615, 0.944, 0.878, and 0.667, respectively. SHapley Additive exPlanations (SHAP) analysis indicated that the Material Resources feature was the most influential predictor for all credits, contributing 20.6% to MRc1.1, 53.4% to MRc2, 36.5% to MRc4, and 22.6% to MRc6. In contrast, the impact of design firms on credit scores was negligible, suggesting that although CDW credits are determined in the design phase, these firms did not significantly influence the decision process. Higher certification levels improved the performance of MRc1.1 and MRc6, while their effect on MRc2 and MRc4 was limited. This study presents a transparent and interpretable XAI-based decision-support framework that reveals the key sustainability drivers of CDW credit performance and provides actionable guidance for LEED consultants, designers, and decision-makers.