Machine learning-based prediction of elliptical double steel columns under compression loading

Abstract This paper presents a comprehensive investigation into the prediction of axial load capacity (P) for elliptical double steel columns (EDSCs) using a diverse set of machine learning models (MLMs). These include Artificial Neural Network (ANN), Gene Expression Programming (GEP), Support Vector Regression (SVR), Random Forest (RF), and AdaBoost. Among the models, AdaBoost demonstrated superior performance, achieving an R 2 of 0.996 and a MAPE of 0.013 during training, outperforming other models under identical conditions. Using a dataset of 119 finite element models derived from prior experimental research, the study validates the proposed solution through k-fold cross-validation, feature importance analysis, and detailed comparisons with experimental data. A Graphical User Interface (GUI) was developed specifically for the AdaBoost model due to its superior accuracy and efficiency, offering engineers a practical and accessible tool for axial load prediction in EDSC design. This research highlights the significance of using advanced machine learning techniques for structural engineering applications, providing valuable insights for the optimization of EDSC performance and design under varying conditions.