Machine learning assisted sustainable mix design of geopolymer concrete: A strength and material efficiency trade-off analysis

Geopolymer concrete is a promising low carbon alternative to ordinary Portland cement concrete; however, its mix design remains challenging because compressive strength is governed by complex interactions among precursor composition, alkaline activator dosage, water demand, aggregate proportion, admixture content, and curing age. This study develops a machine learning assisted framework for evaluating strength material efficiency trade-offs in geopolymer concrete using a literature-based database of 594 mixtures. The original mixture variables were transformed into engineering and sustainability-oriented descriptors, including total binder content, total activator content, water to binder ratio, aggregate to binder ratio, binder efficiency, activator efficiency, and estimated CO₂ and cost indices. An Extra Trees Regressor was selected as the final surrogate model and evaluated using shuffled 10-fold cross-validation, achieving R² = 0.9710 ± 0.0172, MAE = 11.58 ± 2.75 kg/cm², and RMSE = 25.60 ± 9.01 kg/cm². A separate 20% hold-out test used for residual visualization gave R² = 0.9628, MAE = 13.31 kg/cm², and RMSE = 29.88 kg/cm². Statistical and correlation analyses showed high variability in compressive strength, water to binder ratio, binder efficiency, and activator efficiency, suggesting that material performance reflects the combined influence of multiple mixture characteristics rather than individual material quantities alone. Pareto trade-off analysis demonstrated that increasing binder or activator content does not necessarily produce proportional strength improvement. The model predicted balanced optimum mixture achieved a predicted 28-day compressive strength of 595.7 kg/cm² (58.4 MPa), with a total binder content of approximately 370 kg/m³ and a total activator content of 133 kg/m³ . The proposed framework provides a practical data driven tool for preliminary mix design screening and sustainable geopolymer concrete development, while experimental validation remains necessary before practical implementation.