Predicting the eccentric axial load capacity of concrete filled steel tube columns using a machine learning model optimized by Jellyfish Search algorithm

Abstract

Concrete-filled steel tubes (CFST) are becoming increasingly popular in civil and transportation projects due to their significant advantages over traditional reinforced concrete structures. This study aims to develop a predictive model using the Jellyfish Search (JS) optimization algorithm to automatically optimize machine learning parameters for predicting the eccentric compressive strength of CFST. A dataset of 499 samples, comprising 11 input variables and the target variable of
eccentric compressive capacity, was used to develop individual machine learning models (ANN, SVR, CART, LR) and ensemble models (Voting, Bagging, Stacking). After building and comparing these models, the most accurate one was selected for combination with the JS optimization algorithm to achieve the highest-performing predictive model. The results are highly promising, with R = 0.9949, MAE = 46.8157 kN, RMSE = 7.2097 kN, MAPE = 7.67%, and SI = 0.00 (Rank = 1),
indicating that this model holds great potential for CFST design and structural analysis.

Keywords: Concrete filled steel columns; eccentric axial load; machine learning model; optimization algorithm.