On 06/29/2011 09:15 PM, John Hardin wrote: > On Wed, 29 Jun 2011, JKL wrote: > >> mysql> select count(spam_count) from bayes_vars; >> +-------------------+ >> | count(spam_count) | >> +-------------------+ >> | 185 | >> +-------------------+ >> >> mysql> select count(ham_count) from bayes_vars; >> +------------------+ >> | count(ham_count) | >> +------------------+ >> | 185 | >> +------------------+ > > That's not sufficient for Bayes to start scoring messages. The minimum > is 200 each of ham and spam. > > It's generally considered a good idea to train misses and to try to > keep the ratio to something approaching your spam:ham ratio in email > by training regular email that doesn't score really high or low. I > keep mine at about 3:1 spam:ham tokens by training misses and anything > in the 10-80% range. > Agreed. I had been pouring spam into it since Feb. Ham = 1,200 and spam = 400. I still have the original messages so can feed it in again. I have no idea what happened to these data. However, I don't know how to get the data in anymore since its just going to disappear; Its not going into mysql. Not that I mind, because the table is set up for individuals users, and I don't want users mislabelling spam/ham etc. I just want one large database.
Regarding disabling bayes: The only setting in the local.cf is the entry : bayes_auto_learn 0 I read that one has to also comment out yet these are not in the local.cf, but in mysql (not that I think spamass is making use of it). bayes_auto_learn_threshold_nonspam bayes_auto_learn_threshold_spam
