Figure 2
From: Predictors of social risk for post-ischemic stroke reintegration
Predictive modeling framework. (a) Analysis begins with data filtering, using inclusion and exclusion criteria to partition data into the training set for 10-fold cross-validation sampling and hold-out test set for external validation. (b) Model training using GMB methodology tunes hyperparameters during cross-validation and selects the best models for each of the five subsampling methods. (c) Models are validated on the hold-out test set (data not used in model training) to evaluate performance and calculate metrics. (d) Model explainability analysis generates variable importance at the population level, as well as SHAP analysis to identify predictors of social risk at the individual level.
