Enhancing predictive accuracy: a comprehensive study of optimized machine learning models for ultimate load-carrying capacity prediction in SCFST columns

The present study introduces optimized machine learning (OML) models for predicting the ultimate axial load-carrying capacity of square concrete-filled steel tube (SCFST) columns. The structural performance of concrete-filled steel tubular (CFST) members, specifically SCFST columns has gained attention for their superior properties in construction. This study establishes a comprehensive comparative analysis of a hyper-tuned artificial neural network coupled with an improved particle swarm optimization (ANN–IPSO) model to predict the structural behaviour of SCFST columns, considering factors, such as shape, length and height of columns, lateral dimensions of columns, strength of steel and concrete, and thickness of steel tube. The performance of the novel OML model is further compared with conventional algorithms, such as particle swarm optimization (PSO) and grey wolf optimization (GWO). The ANN–IPSO model consistently outperforms other models, demonstrating superior predictive ability and accuracy during both training and validating phases. Furthermore, a novel “Score Analysis” technique is applied to validate the performance of predictive models, showcasing the balanced approach of the IPSO algorithm when coupled with ANN. The study concludes by affirming the consistent superiority of the ANN–IPSO model in predicting the ultimate load-carrying capacity of SCFST columns, expanding the knowledge in engineering studies. The results of the study contribute to engineering knowledge by introducing novel applications of improved machine learning algorithms and emphasising the robustness of the ANN–IPSO model to predict the ultimate load-carrying capacity of columns.