The Area Under an ROC Curve (2001)

The AUC for ROC does not work well in comparisons involving imbalanced datasets, and Precision Recall curves can be a better option.<p><a href="https:&#x2F;&#x2F;classeval.wordpress.com&#x2F;simulation-analysis&#x2F;roc-and-precision-recall-with-imbalanced-datasets&#x2F;" rel="nofollow">https:&#x2F;&#x2F;classeval.wordpress.com&#x2F;simulation-analysis&#x2F;roc-and-...</a>

i would dearly love for this to be a more commonly used measure of performance. it’s much harder to hide your sins if you give a ROC.<p>it’s not naively useful for multi-category classifiers, but it’s great otherwise.

This page actually help me a lot in the past. Thank you for sharing!

I really wish people could make it a habit to define acronyms.