Interpretation of feature impacts on predicted risk in the CoxBoost model using SHAP

The impact of features on the predicted risk in the CoxBoost model was interpreted using SHapley Additive exPlanations (SHAP) values. ​​A: Beeswarm plot of global feature importance, showing the distribution of SHAP values for each feature across all samples, where point color represents the feature value. ​​B: Mean absolute SHAP values for global feature importance ranking. ​​C:SHAP waterfall plot for a single sample, illustrating how each feature contributes to pushing the model's base value (average prediction) higher or lower. ​​D~G: SHAP dependence plots illustrating the relationship between the feature value and its impact on the prediction (SHAP value), with D representing Montreal B classification, E showing C-reactive protein (CRP), F indicating fibrinogen (Fg), and G depicting albumin (ALB).