Open choshin84 opened 5 years ago
PR-curve is more sensitive vs. ROC in imbalanced data set.
https://www.kaggle.com/lct14558/imbalanced-data-why-you-should-not-use-roc-curve
https://qiita.com/ogamiki/items/1ad297ca8761e00e0bca
Further potential evaluation KPI https://pompom168.hatenablog.com/entry/2019/04/11/080000
Tweet summary
PR-curve is more sensitive vs. ROC in imbalanced data set.
Useful link
https://www.kaggle.com/lct14558/imbalanced-data-why-you-should-not-use-roc-curve