A Novel Ensemble Meta-Model for Predicting Demolition Solid Waste Generation

Precise forecasting of Demolition Solid Waste (DSW) generation is essential for the development of sustainable waste management systems. This study uses ensemble machine learning models, such as Random Forest (RF), XGBoost, ANN, and LightGBM, and meta-learners to make predictions more accurate and reliable. The proposed stacked ensemble model shows excellent results, with an R² score of 0.99995 and very low errors in the training, validation, and test datasets, outperforming standalone learners and classic statistical baselines such as SARIMA and ETS. These gains suggest better generalization and stability and lead to practical advantages for operational planning (e.g., capacity sizing, logistics routing, resource allocation, and environmental impact mitigation) in demolition projects. The proposed meta-ensemble model serves as a platform for intelligent real-time decision-support tools, improving strategy selection and system performance in DSW management.