[Spambayes] Untraining?

[G. Armour Van Horn]

I've been running the same instance of SpamBayes (1.1a3) for ages, as a 
proxy, and it's been doing a stellar job. I train this by exception on 
an ongoing basis, classifying all the Unsures and marking any Ham that 
rates over 10% as Ham, and on rare occasions marking something that the 
proxy classified as Ham as Spam. (There have been very few of these to 
deal with.)

However, about three days ago I started seeing a huge increase (maybe 
ten times) in the volume of mail marked as Unsure, all of which seemed 
to be obviously spam to me, along with a dozen or so a day sailing 
through as Ham, often scored as 0.00%.

I recall that this can happen over time, and I'm pretty sure I've solved 
it in the past for others by flushing out all their training and 
starting over, but I can't for the life of me remember how to do that.

The proxy is running on an ancient Fedora Core 5 machine that doesn't 
have much else to do. I see hammie.db and spambayes.messageinfo.db in 
/opt/st_data/ and my database backend is set to dbm.

Van

-- 
----------------------------------------------------------
Sign up now for Quotes of the Day, a handful of quotations
on a theme delivered every morning.
Enlightenment! Daily, for free! 
mailto:twisted at whidbey.com?subject=Subscribe_QOTD

For photography, web design, hosting, and maintenance, 
visit Van's home page: http://www.domainvanhorn.com/van/
-----------------------------------------------------------