Prediction of high-performance concrete strength using machine learning with hierarchical regression

This present study focuses on forecasting the compressive strength (CS) of high-performance concrete (HPC), which is mixed with fly ash and slag, using machine learning techniques with hierarchical quadratic regression (HQ) and multiple linear regression (ML). The study is based on 528 experimental results collected from literature, with variables including cement, blast furnace slag, fly ash, water, super plasticizer, fine aggregate, and coarse aggregate. The primary objective of this study is to develop a model for predicting the CS of HPC. The proposed hierarchical equation outperforms the multiple linear regression model in terms of prediction accuracy. It achieves lower values for mean absolute error (MAE), mean squared error (MSE), root mean square error (RMSE), and mean absolute relative error (MARE), while attaining a higher coefficient of determination (with an R² of 97.12%, adjusted R² of 96.81%, and predicted R² of 96.43%) compared to the ML model (which has an R² of 91.56%, adjusted R² of 91.4%, and predicted R² of 91.17%). A sensitivity analysis was conducted to evaluate the impact of the independent variables on the CS. The results of sensitivity analysis prove that the content of cement significantly impacts the CS of HPC, with a sensitivity analysis parameter of approximately 30%. This is followed by BFS, fly ash, coarse aggregate, fine aggregate, water, and superplasticizer.