Predicting the settlement of pile based on a hybrid form of the model by considering Least Square Support Vector Regression

Understanding the vertical settlement is crucial when designing the pile and foundation type used in real estate, specifically regarding the pile settlement (Sp). This issue is of utmost importance due to the numerous variables involved in designing piles that penetrate rock. Despite multiple efforts, clear and precise theoretical explanations regarding the interactions between soil and piles are currently unclear. As a result, many studies have opted to employ artificial intelligence techniques for determining the subsidence rate of piles over time under different loading conditions. This study presents a machine learning (ML) that effectively predicts the values of Sp, namely Least Square Support Vector Regression (LSSVR). In addition, the proposed model coupled with three meta-heuristic algorithms, including the Flow Direction Algorithm (FDA), Chimp Optimization Algorithm (ChOA), and Rider Optimization Algorithm (ROA), to improve the performance and obtain the optimal results as a framework of hybrid. As a result, LSFD determined the most suitable effects with R2 and RMSE values equal to 0.2503 and 0.9952, respectively. Overall, using LSSVR with FDA, ChOA, and ROA can improve the accuracy and robustness of the model in predicting pile settlement, making it a valuable tool for geotechnical engineers designing foundation systems.