[Spambayes] Spam Bayes Settings
Jesse Pelton
jsp at PKC.com
Wed Aug 2 15:38:22 CEST 2006
To do the best possible job of catching spam, SpamBayes needs to be
well-trained. This means letting it know what you think is spam and what
you think is ham (not spam). If you're using the Outlook plug-in,
training it is simply a matter of using the "Spam" or "Not Spam" button
whenever it fails to classify something correctly.
Once SpamBayes is well-trained, you may find it useful to alter the
cutoffs for "Possible Spam" and/or "Certain Spam." In particular, if you
find that you frequently get messages that are somewhat below the cutoff
for "Certain Spam" and that such messages are never ham, you can lower
that cutoff. It's set fairly high by default to avoid incorrectly
classifying ham as spam. (In the absurd extreme case, you'd set it so
low that all messages were classified as spam. This literally satisfies
your request to "catch the most spam," at the cost of treating your ham
as spam as well, which I doubt is what you want.) If you're using the
Outlook plug-in, you can adjust the thresholds for Possible Spam and
Certain Spam on the Filtering tab of the SpamBayes Manager. Click the
SpamBayes button on the Outlook toolbar, select SpamBayes Manager...
from the menu that pops up, and click the Filtering tab.
There are lots of obscure settings that you can mess with, but unless
you really know what you're doing, you're unlikely to improve the
classification; the defaults are well-chosen.
________________________________
From: spambayes-bounces at python.org [mailto:spambayes-bounces at python.org]
On Behalf Of David Gregory
Sent: Wednesday, August 02, 2006 8:52 AM
To: spambayes at python.org
Subject: [Spambayes] Spam Bayes Settings
What settings should I set SpamBayes at to catch the most spam?
David L. Gregory
Environmental Science Technician
EDWARDS AQUIFER AUTHORITY
1615 North St. Mary's Street
San Antonio, Texas 78215
Phone:(210) 222-2204
Fax:(210) 222-9869
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.python.org/pipermail/spambayes/attachments/20060802/44926d99/attachment.html
More information about the SpamBayes
mailing list