Leakage-safe diffusion augmentation with KAN-based models for imbalanced microarray gene-expression classification

doi:10.1016/j.compbiolchem.2026.109057

Leakage-safe diffusion augmentation with KAN-based models for imbalanced microarray gene-expression classification

2026-04-07

0

PRE

AI

DOI：10.1016/j.compbiolchem.2026.109057

Original

Original request for help

Save

Abstract

En 中文

• Leakage-safe outer cross-validation with real-only validation and testing. • Fold-local preprocessing and hybrid gene selection for microarray gene expression. • Conditional diffusion augments minority classes via anchored sampling and screening. • Macro-F1 is 88.6% on 25 datasets, outperforming Gradient Boosting and XGBoost. • Mismatch checks and fold-stable genes support interpretation in low-support tasks.

Keywords:

Leakage-safe cross-validation

Microarray gene expression

Imbalanced classification

Diffusion augmentation

KAN-based models

Leakage-safe diffusion augmentation with KAN-based models for imbalanced microarray gene-expression classification

Abstract

Journal

Organization

Cited Papers