medium10 pts · first answer countsllm-evaluationmetrics

Macro or micro F1

An intent classifier has one huge class ("order status") and several small but important ones ("fraud report"). Which average better shows how it does on the small classes?

A.Macro F1, which averages each class's F1 equallyB.Micro F1, which pools all predictions togetherC.Accuracy, since it counts every exampleD.Weighted F1, which weights classes by size

Discussion 0

Join free to post in the discussion.

Macro or micro F1

Discussion 0

Related questions