Retour au feed
arXiv cs.LG·

Conformal Risk Prediction for Non-Alcoholic Fatty Liver Disease Using Gradient Boosting with Distribution-Free Coverages

Signal
78
Hype
15
En 3 lignesFramework ML couplant gradient boosting et prédiction conforme pour le risque de NAFLD. Évalué sur 2,599 patients (Guangzhou), atteint AUROC 0.912 en interne et 0.891 en validation externe. Couverture conforme 91.3% à niveau nominal 90%. Stratification 3-niveaux : groupe haut-risque montre taux progression 4.7× supérieur au groupe bas-risque.
Lire la source
Ton avis ?
BenchmarksÉvaluationsSécurité IA

Résumé généré par Claude — vérifié par l'humain