Prediction of compressive strength of blended concrete with Alccofine and GGBFS by applying ensemble machine learning algorithms

Concrete’s physical properties are influenced by the type of supplementary cementitious materials (SCMs) used; therefore, its strength must be assessed before being utilized for specific purposes. The compressive strength of concrete made with ground granulated blast furnace slag and Alccofine 1203 was predicted using ensemble machine learning techniques in this study. While traditional statistical methods have been used for prediction, machine learning approaches offer enhanced accuracy. This study employs ensemble machine learning techniques, including random forest regression and AdaBoost regression, to predict the compressive strength of concrete incorporating ground-granulated blast furnace slag and Alccofine 1203. The models were trained and tested with 160 and 40 experimental conditions, respectively, and their performance was evaluated using MAE, MAPE, MSE, RMSE, SI, R2 and a20ᵢndex. Comparative analysis of prediction accuracy, residuals, and feature importance revealed that while both models performed well, random forest regression outperformed AdaBoost regression, achieving MAE = 0. 69, MAPE = 0. 01, MSE = 0. 87, RMSE = 0. 93 MPa, SI = 0. 02, MBE = -0. 05, R2 = 0. 99 and a20ᵢndex = 1. Feature importance analysis identified “Days” as the most influential factor in both models. These findings confirm that ensemble ML models effectively predict compressive strength, with random forest regression demonstrating superior accuracy.