Application of a support vector regression model employing meta-heuristic algorithms for estimating pile bearing capacity.

The design of pile foundations hinges significantly on the pile-bearing capacity (PBC), a crucial factor affected by numerous soil features and parameters. Piles serve as a key component in relocating structural loads to the ground, underscoring the need for accurately determining PBC in geotechnical design. While prior research has delved into utilizing artificial neural networks (ANN) for predicting PBC, they encounter shortcomings such as grappling with finding global minima and slow convergence rates. This study introduces Support Vector Regression (SVR) as a machine-learning approach, augmented by two meta-heuristic optimization techniques: the Giant Trevally Optimizer (GTO) and Smell Agent Optimization (SAO), to achieve optimal results. These techniques empower engineers and data scientists to develop more precise and reliable predictive models in geotechnical engineering, ultimately enhancing the safety and efficiency of foundation design and construction. Drawing upon an extensive dataset compiled from previous studies, a predictive model was constructed to estimate PBC using soft computing techniques. The outcomes overwhelmingly favored the SVGT model, a fusion of the SVR model with GTO, demonstrating excellent anticipative capacities with an outstanding R2 value of 0.998 and an impressively low RMSE value of 91.21. [ABSTRACT FROM AUTHOR]