Estimation of the axial capacity of high-strength concrete-filled steel tube columns using artificial neural network, random forest, and extreme gradient boosting approaches

The study aims to develop machine learning-based mechanisms that can accurately predict the axial capacity of high-strength concrete-filled steel tube (CFST) columns. Precisely predicting the axial capacity of a CFST column is always challenging for engineers. Using artificial neural networks (ANNs), random forest (RF), and extreme gradient boosting (XG-Boost), a total of 165 experimental data sets were analyzed. The selected input parameters included the steel tensile strength, concrete compressive strength, tube diameter, tube thickness, and column length. The results indicated that the ANN and RF demonstrated a coefficient of determination (R2) value of 0.965 and 0.952 during the training and 0.923 and 0.793 during the testing phase. The most effective technique was the XG-Boost due to its high efficiency, optimizing the gradient boosting, capturing complex patterns, and incorporating regularization to prevent overfitting. The outstanding R2 values of 0.991 and 0.946 during the training and testing were achieved. Due to flexibility in model hyperparameter tuning and customization options, the XG-Boost model demonstrated the lowest values of root mean square error and mean absolute error compared to the other methods. According to the findings, the diameter of CFST columns has the greatest impact on the output, while the column length has the least influence on the ultimate bearing capacity.