Comparative Analysis of Machine Learning Techniques for Prediction of the Compressive Strength of Field Concrete

Abstract

The determination of the concrete compressive strength remains a challenging task in the concrete industry. Machine learning (ML) algorithms offer an alternative and this study presents a comparative analysis of five ML regression models; Gradient Boosting (GB), Random Forest (RF), Decision Tree (DT), K-Nearest Neighbors (KNN), and Linear Regression (LR) on a dataset of 1030 concrete samples. The findings indicate that the GB model achieved the best performance. The developed GB model achieved R-squared values of 91.60%, 91.43%, and 90.18% for the 10-fold, 5-fold, and 3-fold cross-validations, respectively, with mean absolute error, root mean squared error, and mean absolute percentage error values of 2.6776, 4.3523, and 9.19%, respectively. The GB model trained and evaluated was deployed to a web application using Streamlit for real-time prediction of the concrete compressive strength. The results of this research offer a precise and practical method for judging the quality of concrete constructions.