Explainable Predictive Modelling of Sustainable Slag–Fly Ash Based Geopolymer Concrete with Life Cycle and Carbon-Neutrality Assessment

This study proposes a novel, unified framework integrating scientometric analysis, machine learning (ML), explainable artificial intelligence (XAI), and cradle-to-gate life cycle assessment (LCA) to evaluate and predict the performance of slag-fly ash-based geopolymer concrete (SFGPC). A scientometric review of 441 publications (2009–2025) guided the systematic assembly of a dataset comprising 363 SFGPC mixes. Five ML models were trained to predict compressive strength (fc), with Gradient Boosting (GB) achieving the highest accuracy, yielding R2 = 0.954, RMSE = 3.15 MPa, MAE = 1.81 MPa during training, and R2 = 0.95, RMSE = 3.128 MPa, MAE = 2.41 MPa during testing. Multi-layered XAI analysis identified age, slag content, and alkaline-to-binder ratio as the most influential parameters and revealed governing nonlinear interactions. Sustainability assessment showed that the fly ash-dominant mix exhibited the lowest global warming potential (156 kg CO2-eq/m3), the most favourable sustainability index, and the smallest residual emissions after a 25% carbon offset. A user-oriented graphical user interface (GUI) was developed for real-time strength prediction. The novelty of this work lies in introducing an explainable, data-driven, and sustainability-integrated decision-support system for designing transparent and low-carbon geopolymer concretes.