Offshore piles are one of the fundamental components in marine construction projects and are responsible for providing strength and stability to structures against applied loads. Therefore, accurate prediction of the bearing capacity of these piles is of great importance in the design and construction of offshore structures. In this study, the accuracy of different methods for predicting the bearing capacity of steel piles installed in offshore soil deposits of southern Iran is evaluated and compared. For this purpose, in addition to conventional empirical static design methods such as the API approach and cone penetration test (CPT)-based methods including NGI, ICP, UWA, and Fugro, seven machine learning (ML) algorithms were also assessed. These algorithms include Decision Tree (DT), Linear Regression (LR), K-Nearest Neighbors (KNN), Neural Networks (NN), Support Vector Regression (SVR), Light Gradient Boosting Machine (LGBM), and Naïve Bayes (NB).

The results indicate that the Linear Regression (LR) algorithm provides the highest accuracy in predicting the pile tip bearing capacity, followed by the NB algorithm and the API method. Regarding the prediction of pile shaft bearing capacity, the Fugro method demonstrates the best performance, while the NB algorithm and the UWA and API methods rank in the subsequent positions.