Prediction and Interpretability of Axial Load Capacity of Square Concrete-Filled Steel Tube (CFST) Columns Using Machine Learning and Leave-One-Study-Out Validation
Assistant Professor, Department of Civil Engineering, Faculty of Civil Engineering and Architecture, Technical and Vocational University (TVU), Tehran, Iran
10.22124/jcr.2026.33918.1747
Abstract
Concrete-filled steel tubular (CFST) columns are widely used in high-rise buildings and bridges because they combine the high load capacity of steel with the compressive strength of concrete, improve resistance to local buckling, and allow rapid construction. However, predicting their axial strength remains challenging because of the nonlinear composite behavior of steel and confined concrete.This study evaluated three data-driven models—support vector machine with radial basis function kernel (SVM-RBF), multiple linear regression (MLR), and least-squares boosting (LSBoost)—for predicting the axial strength of square CFST columns. A database of 141 experimental specimens was compiled, and the results were compared with predictions from AISC 360-22 provisions.Unlike previous studies that mainly used random K-fold cross-validation, this research employed Leave-One-Study-Out validation. The procedure included 13 iterations, each excluding all specimens from one experimental source during training to assess model generalizability on independent data.The AISC 360-22 provisions produced the highest accuracy, with a correlation coefficient of 0.993. Among the data-driven models, LSBoost performed best, achieving a correlation coefficient of 0.963 and an RMSE approximately 57% lower than that of MLR. SHAP analysis indicated that cross-sectional dimension and steel tube wall thickness were the most influential variables in axial-strength prediction.
Salimbahrami, S. R. (2026). Prediction and Interpretability of Axial Load Capacity of Square Concrete-Filled Steel Tube (CFST) Columns Using Machine Learning and Leave-One-Study-Out Validation. Concrete Research, (), -. doi: 10.22124/jcr.2026.33918.1747
MLA
Salimbahrami, S. R. . "Prediction and Interpretability of Axial Load Capacity of Square Concrete-Filled Steel Tube (CFST) Columns Using Machine Learning and Leave-One-Study-Out Validation", Concrete Research, , , 2026, -. doi: 10.22124/jcr.2026.33918.1747
HARVARD
Salimbahrami, S. R. (2026). 'Prediction and Interpretability of Axial Load Capacity of Square Concrete-Filled Steel Tube (CFST) Columns Using Machine Learning and Leave-One-Study-Out Validation', Concrete Research, (), pp. -. doi: 10.22124/jcr.2026.33918.1747
CHICAGO
S. R. Salimbahrami, "Prediction and Interpretability of Axial Load Capacity of Square Concrete-Filled Steel Tube (CFST) Columns Using Machine Learning and Leave-One-Study-Out Validation," Concrete Research, (2026): -, doi: 10.22124/jcr.2026.33918.1747
VANCOUVER
Salimbahrami, S. R. Prediction and Interpretability of Axial Load Capacity of Square Concrete-Filled Steel Tube (CFST) Columns Using Machine Learning and Leave-One-Study-Out Validation. Concrete Research, 2026; (): -. doi: 10.22124/jcr.2026.33918.1747
)