http://bugzilla.spamassassin.org/show_bug.cgi?id=2853





------- Additional Comments From [EMAIL PROTECTED]  2004-05-14 13:01 -------
Subject: Re:  Rewrite masses/ (in perl)

> > I'm almost done on this. I'd like to see it in 3.0.0. I need to do a
> > little more testing and clean up the documenation, but my (quick) results
> > suggest an average TCR of 64.622 (scoreset 2) which is better than the old
> > stuff which had a TCR of 56.288. (I did make a bit of a change to the
> > algorithm to score ranges to use soratio instead of rank, so that probably
> > accounts for much of the difference.)
> 
> Cool!  one thing though -- I found that relying entirely on soratio
> could be bad news when the hit-rates are too small; e.g.
> a test that hits 0.05% of spam and 0% ham may get a score of 4.0.
> This could be tricky if the training corpus isn't representative
> of other corpora, which might have some ham hits.

Those numbers are based on 10FCV.




------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

Reply via email to