On Tue, Jan 14, 2003 at 06:45:32PM -0500, Yevgeniy Miretskiy wrote: > OVERALL% SPAM% HAM% S/O RANK SCORE NAME > 4451 1900 2551 0.427 0.00 0.00 (all messages) > 100.000 42.6870 57.3130 0.427 0.00 0.00 (all messages as %) > > What should I look for to determine whether a particular rule is good > or bad? What is a highest possible score? What does rank of 1.0 > mean?
It's more of an art than a science. Typically I would look at match %,
S/O, and RANK. If it looks right (high match %, S/O near 0 or 1 depending
on your goal (match ham or spam respectively), and RANK should be > .9.
It's more of a feeling than hard limits.
Highest possible score? With mass-check, the scores are the ones set
in the cf files. It doesn't generate scores.
A rank of 1 means the rule is considered to be very good for your corpus.
--
Randomly Generated Tagline:
"Speech is conveniently located midway between thought and action, where
it often substitutes for both."
- jACL on Slashdot, comment #4594955
msg12197/pgp00000.pgp
Description: PGP signature
