Machine learning-driven optimization for predicting compressive strength in fly ash geopolymer concrete

This study employs machine learning (ML) techniques to predict the compressive strength ( f c ′ ) of fly ash-based geopolymer concrete, utilizing a comprehensive set of experimental data. The analysis considered variables such as fly ash (FA) components, coarse and fine aggregates, alkaline activator molarity, and other additives. Six different ML models—AdaBoost, Decision Tree, Extra Tree, Random Forest, Gradient Boosting, and Extreme Gradient Boosting were used to predict f c ′ of fly ash-based geopolymer concrete. The results reveal that the AdaBoost model outperformed the other models, achieving R 2 score of 0.80 and RMSE of 6.60. Furthermore, the tuned models demonstrated superior accuracy compared to their default counterparts. The feature importance analysis using the Shapley values technique identified CaO as the most influential factor on f c ′ , with higher CaO levels leading to an increase in compressive strength. Additionally, an increase in the molarity of the NaOH alkaline activator positively impacted the target value.