XGBoost-Based Prediction of the 28-Day Compressive Strength of Portland Cement Concrete Containing Blast Furnace Slag and Fly Ash

Reducing Portland Cement (PC) consumption through Fly Ash (FA) and Blast Furnace Slag (BFS) can lower concrete’s environmental burden, but the coupled effects of these constituents make 28-day compressive strength difficult to estimate with conventional empirical relationships. Accordingly, this study developed an interpretable Extreme Gradient Boosting (XGBoost)-based model for predicting the 28-day compressive strength of high-performance Portland cement concrete incorporating these supplementary cementitious materials. For this purpose, a database comprising 415 normally cured mixtures tested at 28 days was adopted. Eight predictors were considered: theoretical mixture density and the gravimetric proportions of PC, FA, BFS, water, superplasticizer, coarse aggregate, and fine aggregate. The predictive performance was assessed using shuffled six-fold cross-validation. Meanwhile, pooled out-of-fold predictions achieved a coefficient of determination (R2) of 0.85, a mean absolute error of 3.55 MPa, and a root mean squared error of 5.62 MPa. In addition, fold-specific R2 values ranged from 0.80 to 0.90. Native XGBoost importance and SHapley Additive exPlanations ranked PC content, water content, and theoretical density as the leading input variables in terms of predictive significance. In this way, the proposed XGBoost model can support preliminary strength estimation and screening of eco-friendly mix designs before laboratory verification. Regardless, its use should remain within the represented compositional ranges until independent validation confirms transferability.