Concrete-filled steel tube (CFST) columns are composite structural members that have gained significant attention in modern civil engineering due to their high strength, ductility, and efficient interaction between steel confinement and concrete compression. However, the complex nonlinear behavior governing their load-carrying capacity makes accurate prediction challenging when using conventional design approaches. In this study, data-driven artificial intelligence techniques, including Multivariate Adaptive Regression Splines (MARS) and Artificial Neural Networks (ANNs), were developed to estimate the load-carrying capacity of CFST columns. To enhance predictive performance and overcome the limitations of traditional parameter calibration, the Water Cycle Algorithm (WCA) was integrated as a metaheuristic optimization strategy. The hybrid intelligent–metaheuristic framework demonstrated improved predictive accuracy and computational efficiency. The correlation coefficients obtained during the training phase for the standalone MARS and WCA-optimized MARS models were 0.991 and 0.969, respectively, with corresponding RMSE values of 217.19 KN and 298.21 KN. In addition, comparisons with established design provisions, including ACI 318, LRFD, and Eurocode 4, confirmed the superior performance of the proposed data-driven models in predicting CFST load-carrying capacity. Sensitivity analysis further revealed that section diameter, wall thickness, and member length are the most influential parameters, with contribution ratios of 38.16%, 23.05%, and 17.19%, respectively.
Abbasi et al. (Wed,) studied this question.