1
$\begingroup$

I have a question with regard to the proper way to report F1 scores. Say I am comparing two algorithms one with F1 score of 0.71 and the other of 0.82.

Is it correct to say:

"Algorithm 1 obtained an F1 score 11 points higher than algorithm 2"

or

"Algorithm 1 obtained an F1 score 11 percentage points higher than algorithm 2"

or

"Algorithm 1 obtained an F1 score 0.11 points higher than algorithm 2".

Or none of these? Some other way? I suppose a second question for this is if it is correct to report the scores as 0.71 and 0.82 or more correct to say 71% and 82%.

$\endgroup$
2
$\begingroup$

I would certainly not write about "11 points higher". The F1 score is a number between 0 and 1 and can be interpreted as a percentage, so "11 percentage points" is defensible, but it's certainly not standard to refer to 0.01 as "1 point".

I believe everyone will be quite as comfortable with "71%" as with "0.71".

And of course the F1 (or F$\beta$) score suffers from all the same problems that accuracy as an evaluation metric has.

$\endgroup$
  • $\begingroup$ I knew someone was going to say something about F1 score being improper... people just can’t help themselves $\endgroup$ – astel Sep 10 at 19:31
0
$\begingroup$

Yes, you can interpret F1 score as 71% in one model and 82% in another model. Calling them as higher or lower than another model is not the correct way.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.