I have posted similar question before...
I want to find out ways to erase what the Bayesian's database have learned. The reason for that because I messed it up when I first install SA. I trained it with too many spams and little hams, instead of 200/200 each.
Is it actualy causing problems? This, by itself, is not a problem. Don't think you need to train an equal number of each, just because the minimums are both 200. This has been discussed *many* times on the list. It's a common misconception, but a completely false one.
If you don't beleive me, look at my training ratio: 0.000 0 276257 0 non-token data: nspam 0.000 0 10219 0 non-token data: nham
That's a 27:1 spam:ham ratio... No problems here.
I have tried using sa-learn --rebuild, and sa-learn --force-expire. Is there a way to do it instead of reinstall SA?
Even reinstalling won't wipe your bayes database, since the database lives in the user's home directory. You'll have to delete the database directly.
rm ~/.spamassassin/bayes_*
