[Spambayes] Spambayes works so well, it's hard to keep training balanced

skip at pobox.com skip at pobox.com
Tue Oct 17 13:25:10 CEST 2006


    Claude> Over time, the result is that I've built up a huge imbalance of
    Claude> trained messages, nearly 1000 trained spam vs. 150 trained ham

    Claude> So, how to regain balance?

Are your hams and spams stored in a plain old mbox file or something your
mail reader can read?  If so, visit your saved spam, sort by date and delete
a bunch of the oldest ones (maybe a couple hundred to start with).  I'm
currently running with a 4:1 ratio using train-to-exhaustion without any
real problems.

Skip


More information about the SpamBayes mailing list