| Abstract Scope |
Magnesium (Mg) alloys with controlled biodegradation are attracting increasing attention for use in temporary implant applications. However, optimizing their mechanical performance remains challenging due to the coupled effects of alloy chemistry and processing history. In this study, a machine learning–based strategy is introduced to support the accelerated development of diluted Mg alloys for biomedical applications. A comprehensive dataset was employed to train and evaluate six predictive models targeting mechanical properties. Among the ML approaches, ensemble learning methods showed superior performance, with the CatBoost algorithm providing the most accurate predictions. To enhance physical interpretability, SHAP were applied, highlighting the dominant influence of thermomechanical processing routes and specific alloying additions, particularly Zn, Mn, and Gd. The predictive capability of the models was validated against experimental data, demonstrating reliable generalization. Finally, the trained model was used to construct composition-dependent property maps, enabling visualization of strength–ductility trade-offs in Zn–Mn-containing Mg alloys. |