Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. It only takes a minute to sign up.
Sign up to join this community
In a multi-label or multi-class classification setting, when choosing between a micro or a weighted F1 score, what shall I take into account?
The main upside of choosing macro is that one gets a sense of effectiveness on small classes. Assuming that the small classes are of no importance, how to choose between micro or a weighted F1 score?
Micro f1 is based on global precision and recall. It treats each test case equally and doesn't give advantages to small classes. I think it's more suitable.
This article "Macro- and micro-averaged evaluation measures" from Vincent Van Asch in U of Antwerp explains many different kinds of f1 scores.
