GBMP-LCA: A Gradient Boosting–Based Framework for Performance Prediction and Life-Cycle Optimization of Green Building Materials

Unofficial AI-generated summary based on the public title and abstract. Not an official translation.

🔬研究者:Provides a novel application of gradient boosting and SHAP for multi-objective material optimization, enabling interpretable ML in LCA.

🏢実務担当者:Offers a practical tool for screening and optimizing green building materials based on performance and carbon footprint.

🏛政策担当者:Supports evidence-based policy for low-carbon construction by quantifying trade-offs between durability and emissions.

The accelerated adoption of green building materials is essential for reducing the environmental footprint of the construction sector, yet accurately predicting their long-term performance and lifecycle impacts remains a major challenge due to complex material compositions and heterogeneous durability mechanisms. Conventional experimental and physics-based approaches are often costly and limited in scalability, motivating the integration of data-driven methods for sustainable material design and optimization. In this study, we propose GBMP-LCA, a Gradient Boosting–Based framework for performance prediction and life-cycle optimization of green building materials. The proposed framework systematically integrates material composition parameters, physical and mechanical properties, durability indicators, and embodied carbon metrics to model key performance outcomes such as service life and strength retention. A LightGBM-based gradient boosting model is employed to capture nonlinear relationships among multi-source features, while SHAP-based interpretability analysis is introduced to quantify feature contributions and guide material composition optimization under sustainability constraints. Experimental results on a compiled dataset of diverse green building materials show that the proposed GBMP-LCA framework consistently outperforms conventional machine learning baselines. Compared with Random Forest and linear regression models, GBMP-LCA reduces RMSE by approximately 12% and achieves an  value of 0.87, indicating strong predictive accuracy and robustness. Furthermore, model interpretability analysis reveals key trade-offs between material durability improvement and carbon footprint reduction, supporting data-driven sustainable material selection. Overall, GBMP-LCA offers an accurate, explainable, and scalable decision-support framework for sustainable building material development, with practical applicability to material screening, lifecycle-aware design, and low-carbon construction planning.

• semanticscholar https://fsrma.org/index.php/FSRMA/article/download/105/108first seen 2026-07-13 07:18:08