Table 1.
Summary of baseline characteristics in breast cancer patients, stratified by survival status (censored vs deceased).
Fig 1.
Kaplan-Meier survival curves of demographic factors on the overall dataset ( N = 4 , 024 ) of breast cancer patients with pairwise comparisons.
(A) KM curve for age groups and the associated risk table. (B) KM curve for race and associated risk table. (C) KM curve for marital status and associated risk table.
Fig 2.
Kaplan-Meier survival curves of clinical factors on the overall dataset ( N = 4 , 024 ) of breast cancer patients with pairwise comparisons.
(A) KM for N stage and associated risk table. (B) KM for T stage and associated risk table. (C) KM curve for tumor size and associated risk table.
Fig 3.
Feature importance and beeswarm plot determined by mean absolute SHAP values for the random survival forest (RSF) model.
(A) SHAP value distribution for features, highlighting the relationship between feature values and their impact on model predictions. (B) Bar plot of mean absolute SHAP values by decreasing order of importance.
Table 2.
Results of the multivariate analysis of the CPH model after stratification for key variables.
Fig 4.
Feature importance and beeswarm plot determined by mean absolute SHAP values for the DeepSurv model.
(A) A comprehensive view of the influence each variable has on model predictions, with N Stage Status ranked on top (B) Bar plot of mean absolute SHAP values by decreasing order of importance
Fig 5.
Bee-swarm visualization, sorted by the mean absolute SHAP values, showcasing the variables’ influence on the models’ predictions.
(A) SHAP feature Importance of the XGBoost model. (B) SHAP feature Importance of the RF model. (C) Comparison of ROC-AUC Scores of the ML models.
Table 3.
Comparison of models employed with their respective performance measures.