Machine learning models for predicting the axial compression capacity of cold‑formed steel elliptical hollow section columns

This study presents the performance of three machine learning (ML) models including gradient boosting regression trees (GBRT), artificial neural network model (ANN), and artificial neural network–particle swarm optimization (ANN-PSO) for predicting the axial compression capacity (ACC) of cold‑formed steel elliptical hollow section (EHS) columns. To achieve the goal, a set of 291 data is collected from previous studies to develop GBRT, ANN, and ANN-PSO models. The performance of GBRT, ANN, and ANN-PSO models is evaluated based on the statistical indicators, which are R2,RMSE,MAPE,and i20-index. The results show that the ANN-PSO model with R2=1.00,RMSE=41.3631,MAPE=1.3689,and i20-index=0.9966has the best performance compared to GBRT and ANN models. Moreover, a graphical user interface tool is developed based on the ANN-PSO model for practical designs.