Prediction and optimization of strength and CO2 emission for geopolymer concrete mix design using machine learning

The growing demand for sustainable construction materials has driven interest in geopolymer concrete (GPC) as an eco-friendly alternative to conventional Ordinary Portland Cement (OPC) concrete. GPC offers considerable potential to reduce carbon emissions. However, the lack of standardized mix design procedures and the complex interdependence of its components pose significant challenges in achieving optimal performance. This study presents a machine learning (ML)-based approach for optimizing GPC mix design with the dual objective of minimizing CO 2 emissions and maximizing compressive strength. A comprehensive dataset comprising 1507 data points was compiled from 23 published studies. The Extreme Gradient Boosting (XGBoost) model was employed and achieved strong predictive performance with a Coefficient of Determination (R 2 ) of 0.9113, Mean Absolute Error (MAE) of 3.9232, and Root Mean Squared Error (RMSE) of 6.2736. Further, Bayesian Optimization (BO) was applied to identify optimal mix proportions that meet strength requirements while minimizing environmental impact. The optimization results a 35.4 % increase in compressive strength and a 77.7 % reduction in carbon emissions. Additionally, a Graphical User Interface (GUI) was developed to allow users to input their mix design parameters, receive instant predictions and optimization recommendations. This study highlights the potential of integrating ML with sustainable construction practices and offers a scalable, data-driven solution for environmentally responsible concrete design. • XGBoost model predicted GPC compressive strength with an R 2 of 0.91. • Bayesian Optimization improved strength by 35.4 % and cut CO 2 by 77.7 %. • 1507 data points compiled from 23 studies ensured robust model training. • A GUI tool was developed for instant GPC mix prediction and optimization. • The model integrates sustainability by optimizing CO 2 and strength trade-offs.