subject-tagging test results
Greg Louis
glouis at dynamicro.on.ca
Fri Feb 14 14:13:41 CET 2003
On 20030214 (Fri) at 0806:39 -0500, Greg Louis wrote:
> I ran a test...
> Speaking of parameters, Robinson's x in this experiment was 0.415 and s
> was 5.0e-7. The spam cutoffs determined from the nonspam test corpora
> were quite high, ranging from 0.991 to 0.998; that reflects the fact
> that the nonspam corpora included quite a few spammy-looking
> newsletters, so the high cutoffs are needed to avoid false positives.
> Even so, we caught up to 96.5% of spam -- not that bad a result.
Forgot (because I never use any other method) to mention that this
experiment was done with the Robinson-Fisher classification algorithm.
--
| G r e g L o u i s | gpg public key: |
| http://www.bgl.nu/~glouis | finger greg at bgl.nu |
| Help free our mailboxes. Include |
| http://wecanstopspam.org in your signature. |
More information about the Bogofilter
mailing list