2 Different F1-Measure to calculate clustering performance - which one is correct and why?

I know it sounds incorrect but that is the truth

Here let me show you

This below one is the first one and very widely used in the literature First one reference : Steinbach, Michael, George Karypis, and Vipin Kumar. "A comparison of document clustering techniques." KDD workshop on text mining. Vol. 400. No. 1. 2000.

And the second one is this one : http://nlp.stanford.edu/IR-book/html/htmledition/evaluation-of-clustering-1.html

Here an answer about second one : https://stats.stackexchange.com/a/80194/5263

The second one is pretty complex and hard to understand

I coded both and they produce different results

I believe second one is better evaluation but the question is

Why there are 2 different F1-Measure? Which one is better and why?

• At the end of your first page, it says $F_\beta = \frac{(1+\beta^2)PR}{\beta^2P + R}$, which, for $\beta = 1$, is the $F_1$ score. This is the $F_\beta$ score, a generalization (see Wikipedia). What did you mistake for the $F_1$ score in your link? – Winks Mar 30 '16 at 21:37
• @Winks so you say F1-Score and F-Measure are different things? – MonsterMMORPG Mar 30 '16 at 21:48
• @Winks because wikipedia tells F-Measure and F1-Score are same things – MonsterMMORPG Mar 30 '16 at 21:48
• @Winks 1 more thing. How do you format your answer that way in comment section :D – MonsterMMORPG Mar 30 '16 at 21:51