# GRASP: group-Shapley feature selection for patients

arXiv: https://arxiv.org/abs/2602.11084 · HTML: https://mlsanity.com/paper/2602.11084/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 8 of its 22 references that are in the map.
- First posted on arXiv 11 Feb 2026.

## Builds on (most cited first, 8 of 8)

- [A Unified Approach to Interpreting Model Predictions](https://mlsanity.com/paper/1705.07874/) (2017)
- [XGBoost: A Scalable Tree Boosting System](https://mlsanity.com/paper/1603.02754/) (2016)
- [Optuna: A Next-generation Hyperparameter Optimization Framework](https://mlsanity.com/paper/1907.10902/) (2019)
- [A Review of Challenges and Opportunities in Machine Learning for Health](https://mlsanity.com/paper/1806.00388/) (2018)
- [Feature Selection: A Data Perspective](https://mlsanity.com/paper/1601.07996/) (2016)
- [AFS: An Attention-based mechanism for Supervised Feature Selection](https://mlsanity.com/paper/1902.11074/) (2019)
- [An exhaustive variable selection study for linear models of soundscape emotions: rankings and Gibbs analysis](https://mlsanity.com/paper/2207.12743/) (2022)
- [Enhancing Graphical Lasso: A Robust Scheme for Non-Stationary Mean Data](https://mlsanity.com/paper/2503.19651/) (2025)

## Built on by (most cited first, 1 of 1)

- [LUNG-KGMM: Knowledge-Guided Multimodal Learning for Lung Cancer Incidence Prediction](https://mlsanity.com/paper/2608.14657/) (2026)

## Abstract (arXiv)

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 L21 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 L21 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.