• Developed an interpretable SSA-XGBoost model achieving high accuracy (R²=0.91) for predicting stretch formability (IE) of AZ31 Mg alloys.
• Identified ten key input features from microstructure, mechanical properties, and test conditions, with Imax, TYS, EL, r, GS, and ΔS as most influential.
• Validated model generalization with new experimental data, showing prediction errors below 5%.
• Provided quantitative insights via SHAP analysis, aiding the design of high-formability magnesium alloys.