Feature selection methods compared for opioid use disorder prediction

Researchers have compared five feature selection methods for predicting opioid use disorder (OUD) using electronic health records (EHR). The study evaluated recurrence enrichment, NTK-motivated early gradient sensitivity, LightGBM-SHAP, Elastic Net, and LLM-guided semantic selection. NTK sensitivity offered the best balance of accuracy and stability, while LLM-guided selection provided complementary clinical insights. AI

IMPACT This research could improve the accuracy and interpretability of AI models used in clinical settings for predicting opioid use disorder.

RANK_REASON The cluster contains an academic paper detailing a comparative study of methods. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Elastic Net
• electronic health records
• large language model
• LightGBM-SHAP
• NTK-motivated early gradient sensitivity
• opioid use disorder
• recurrence enrichment

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →