Sustainable optimisation of GGBS-based concrete: De-risking mix design through predictive machine learning models

Ground Granulated Blast Furnace Slag (GGBS) is increasingly recognised as a sustainable alternative to traditional Portland cement in concrete. However, predicting the compressive strength (C-S) of GGBS-based mixes remains challenging due to complex material interactions. This study applies four supervised machine learning (ML) algorithms, Decision Tree, Random Forest, Gradient Boosting, and XGBoost, to predict the C-S using a literature-derived dataset. Among these, XGBoost exhibited the best performance (R² = 0.979) with the lowest prediction error. SHAP analysis reveals that cement content, curing age, and water-to-binder ratio are the most influential features. To enhance practical utility, a graphical user interface (GUI) was developed for real-time strength prediction based on user-defined input parameters. The proposed framework demonstrates the potential of ML to support accurate, efficient, and sustainable mix design in real-world construction scenarios.