Predicting pile-bearing capacity utilizing least square support vector regression coupled with giant trevally optimizer and the flying foxes optimization

Addressing the intricacies of pile analysis and design poses a complex challenge, particularly in leveraging machine learning (ML) approaches to predict the ultimate load-bearing capacity of piles based on field trial data. This study focuses on developing sophisticated ML prediction models tailored for calculating pile-bearing capacity (Pu). The foundation of this prediction model relies on the least square support vector regression (LSSVR) technique. A distinctive hybrid strategy is employed, combining the giant trevally optimizer (GTO) and flying foxes optimization (FFO) approaches. This unique amalgamation ensures precise and optimal prediction outcomes. The research relies on a meticulously curated dataset encompassing a diverse array of parameters related to piles and soil properties sourced from various literature. These datasets form the robust foundation for training, validating, and testing the models developed in this study. The implemented methodology has yielded remarkably accurate results, affirming the efficacy of the proposed models. This integration has led to the emergence of three hybrid models: LSFF, LSGT, and LSSVR. Particularly noteworthy is the exceptional performance of the LSFF model, demonstrating outstanding accuracy with an impressive R² value of 0.993 and a minimal RMSE value of 32.916. These metrics underscore a notable achievement in the precise estimation of Pu, establishing the reliability and effectiveness of the approach taken in this research. The unique hybrid strategy, informed by diverse parameters, improves predictive capabilities, ensuring safer and more robust infrastructure projects by precisely understanding load-bearing capacities in different soil conditions.