A novel XGBoost-based model to accurately predict envelope and modeling parameters for rectangular RC columns

Abstract

ASCE 41-17 provides empirical equations to estimate the envelope and nonlinear modeling parameters for reinforced concrete (RC) columns. It is, however, a difficult task to accurately compute the envelope and modeling parameters using empirical equations obtained from regression analyses or conventional machine learning models because RC columns behave in a highly nonlinear manner in their inelastic range and the relationships between input and output parameters (envelope and modeling parameters) are complex. To address such difficulties, we propose a novel XGBoost (eXtreme Gradient Boosting)-based model that can accurately predict the envelope and modeling parameters for RC columns. A database containing the test data of 275 rectangular RC columns was constructed to develop the model. The accuracy of the proposed model is verified by conducting cross validation and comparing the results with those obtained from ASCE 41-17 empirical equations and a conventional XGBoost model.