Waste-Derived Geopolymer Mortars from Industrial Slags: Modeling Early-Age Compressive Strength for Sustainable Construction

This study aims to predict the compressive strength of mortars produced with GBFS and SMS. A total of 72 samples were prepared using six mixture ratios and four Na₂SiO₃ dosages and cured at 75 °C for 24 h. The dataset was analyzed using multivariate regression and machine learning models. Among regression approaches, the power function performed best (RMSE ≈ 4.9204 MPa, NSE ≈ 0.9284). However, the Gaussian Process (GP) model outperformed all other methodologies with the lowest prediction error and exceptional correlation (RMSE ≈ 1.4203 MPa, NSE ≈ 0.9920).