<p>Drag Embedment Anchors (DEAs) are widely utilized in offshore engineering for mooring systems due to their cost-efficiency and scalability. Accurate prediction of DEA penetration depth and holding capacity is critical for optimal design and installation. Traditional methods, such as empirical models, finite-element simulations, and field experiments, are often time-consuming, expensive, and limited in scope. This study explored the application of Machine Learning (ML) techniques, e.g., Artificial Neural Networks (ANN), Support Vector Regression (SVR), and Extra Tree Regression (ETR), to model DEA behavior in sand seabed. A comprehensive dataset derived from experimental and analytical studies was used to train and validate the models. <i>K</i>-fold cross-validation (<i>K</i> = 10) was employed for all ML models (ANN, SVR, and ETR) in the current study. Key input parameters, including anchor geometry, soil properties, and installation conditions, were utilized to develop ML models with different combinations of inputs. Among the models, ETR demonstrated superior performance in predicting holding capacity and penetration depth. Sensitivity analysis revealed that soil parameters, such as friction angle (<i>φ</i>) and dilation angle (<i>ψ</i>), along with anchor weight, were critical for accurate predictions, whereas geometric ratios played a secondary role. ETR’s ensemble approach effectively captured complex feature interactions, making it the most robust model for holding capacity estimation. The findings highlighted the potential of ML techniques to complement traditional methods, offering faster and more accurate predictions for DEA design optimization and offshore installation planning. This study underscored the importance of integrating geotechnical knowledge into ML frameworks to enhance interpretability and practical relevance in offshore engineering applications.</p> Graphical abstract <p></p>