Predicting Mechanical Strength of Alkali-Activated High-Performance Concrete Using Machine-Learning Methods

The growing demand for concrete poses a significant environmental challenge, but alkali-activated high-performance concrete (AA-HPC) offers a more sustainable alternative by potentially reducing carbon emissions and ecological harm. This study explores the latest developments in machine learning (ML) applications aimed at predicting the compressive strength of AA-HPC, with a focus on minimizing experimental expenses, construction duration, and environmental impact. Among nine evaluated ML models, the combination of extreme gradient boosting (XGBoost) with the African vultures optimization algorithm (AVOA) emerged as the most effective. AVOA proved highly efficient in optimizing model parameters, achieving the lowest root mean square error (RMSE) during hyperparameter tuning. On the training dataset, XGB-AVOA reached an R² of 0.994 and an RMSE of 2.368, while on the testing dataset, it maintained superior performance with an R² of 0.975 and an RMSE of 5.664. These findings highlight AVOA's strength in fine-tuning XGBoost compared to alternative optimizers such as grey wolf optimizer (GWO), whale optimization algorithm (WOA), social spider optimization (SSO), and gorilla troops optimizer (GTO). To support practical implementation, a graphical user interface (GUI) has also been developed, allowing researchers to efficiently utilize the XGB-AVOA model for accurate, cost-effective, and time-saving predictions in laboratory environments.