Data leakage inflates prediction performance in connectome-based machine learning models
Sign Up to like & getrecommendations! Published in 2024 at "Nature Communications"
DOI: 10.1038/s41467-024-46150-w
Abstract: Predictive modeling is a central technique in neuroimaging to identify brain-behavior relationships and test their generalizability to unseen data. However, data leakage undermines the validity of predictive models by breaching the separation between training and… read more here.
Keywords: leakage; prediction performance; inflates prediction; machine learning ... See more keywords