Payal Rathod <[EMAIL PROTECTED]> wrote on 10/04/2004 12:54:46 PM:
> On Mon, Oct 04, 2004 at 12:48:18PM -0500, Andy Jezierski wrote:
> > Yep, bayes is working, notice the BAYESS_99 rule hit above. Another
thing
> > to keep in mind is that Bayes only kicks in after you have 200 ham &
200
> > spam messages learned. Looks like your bayes has only recently gone
above
> > the 200 mark.
>
> Thanks a lot for the mails all of you. For how many more messages do I
have
> to train bayes? It is rather a tiring job.
>
> -Payal
>
You have to continually keep bayes updated! If not, it won't help much
with the ever changing nature of spam messages. Read through the docs
about bayes auto-learning to see if it's something that you might want to
implement. Some people use it, others don't. It's a matter of personal
preference as well as how well you trust your existing SA implementation to
correctly classify incoming messages. If you are currently getting a lot
of false positives or false negatives, I'd think twice before turning on
auto learning. If you pollute your bayes database with incorrectly
classified messages, your users will start screaming, and you'll have to
trash your bayes database and start all over again.
Andy
