Prediction of Concrete Compressive Strength Using Boosting-Based Machine Learning Algorithms

초록

This study investigates the use of boosting-based machine learning algorithms to predict the compressive strength of concrete, aiming to improve structural safety, optimize mix proportions, and enhance construction efficiency. Traditional empirical models often fall short in modeling the complex nonlinear relationships among input materials. Five popular boosting methods—AdaBoost, Gradient Boosting Machine (GBM), XGBoost, LightGBM, and CatBoost—were evaluated using a benchmark dataset of 1030 samples from the UCI repository, containing eight numerical features. Model performance was measured using the coefficient of determination (R2). Among the methods, CatBoost outperformed others with R2 = 0.9943 on the training set and 0.9440 on the testing set, followed by XGBoost and GBM. AdaBoost showed the weakest performance. The results highlight the strong capability of advanced gradient boosting algorithms, particularly CatBoost, in modeling the nonlinear behavior of concrete materials. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.

키워드