"starter bayes" database

[hvdkooij at vanderkooij.org]

Jan Agermose wrote:

> 2) is it possible to download and install a “base bayes database”? The
> result of training the database from some sort of standard accepted
> training set of spam/ham mail?

No. Each network is unique and you must train your bayesian database to
fit YOUR traffic and not fit someone elses traffic.

I would even recommend against autolearning. If you manually select 200
SPAM and 200 HAM messages you make the database work for you. In this
regard read the instructions from the competition:
http://www.barracudanetworks.com/ns/downloads/Barracuda_Bayes.pdf

If you do it this way in MailScanner you get pretty good results. I
actually disable autolearn and use mailwatch to train the database. For
that you need to store all messages in quarantaine.

Hugo.

-- 
hvdkooij at vanderkooij.org               http://hugo.vanderkooij.org/
	Don't meddle in the affairs of sysadmins,
	for they are subtle and quick to anger.