A QSAR classifier is trained for virtual screening on a dataset with very few active compounds. Which evaluation concern is most important?

Question

Answer

Class imbalance may make precision and recall unstable, so evaluation should rely only on AUC-ROC.

Answer

Class imbalance may make accuracy too pessimistic, so overall error rate should be preferred over active-class recovery.

Answer

High overall accuracy may hide poor recovery of active compounds, so metrics such as precision, recall, AUC-PR, MCC, or early-recognition metrics should be examined.

Crowdly

A QSAR classifier is trained for virtual screening on a dataset with very few ac...

Want instant access to all verified answers on moodle.jku.at?