GRASP: group-Shapley feature selection for patients

TL;DR

GRASP is a new feature selection framework for medical prediction that combines Shapley value attribution with group regularization to produce stable, interpretable, and non-redundant feature sets.

Contribution

It introduces a novel combination of SHAP-based importance scoring with group L21 regularization for robust feature selection in medical prediction.

Findings

GRASP achieves comparable or better accuracy than existing methods.

It identifies fewer, more stable, and less redundant features.

GRASP outperforms LASSO, SHAP, and deep learning methods in experiments.

Abstract

Feature selection remains a major challenge in medical prediction, where existing approaches such as LASSO often lack robustness and interpretability. We introduce GRASP, a novel framework that couples Shapley value driven attribution with group $L_{21}$ regularization to extract compact and non-redundant feature sets. GRASP first distills group level importance scores from a pretrained tree model via SHAP, then enforces structured sparsity through group $L_{21}$ regularized logistic regression, yielding stable and interpretable selections. Extensive comparisons with LASSO, SHAP, and deep learning based methods show that GRASP consistently delivers comparable or superior predictive accuracy, while identifying fewer, less redundant, and more stable features.