You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@spamassassin.apache.org by Khalid Waheed <kh...@eim.ae> on 2004/09/28 11:05:51 UTC

Bayes filter

Dear all,

What is the importance of bayes filter in spamassassin? it takes a lot of CPU resource so I have disabled it and have not

felt much difference.

Thanks,

Kha



Re: Bayes filter

Posted by Erik Wickstrom <er...@gmail.com>.
If this is a new install, you won't feel much of a difference.  Bayes
however is an extremely powerful tool once it is trained.  Bayes
"learns" what your ham and spam emails look like, and then through the
magic of statistics, it classifies incoming messages as spam or ham
based on your history.  I would personally recommend not disabling it
- but do take the time to train it.  (note training takes a minumum of
200 spams and 200 hams - although to maximize effectivness you will
want to use up to 5000 of each.  If you don't have that many, don't
worry, Bayes will also auto learn....)

Give it a try!

Erik


On Tue, 28 Sep 2004 13:05:51 +0400, Khalid Waheed <kh...@eim.ae> wrote:
> Dear all,
> 
> What is the importance of bayes filter in spamassassin? it takes a lot of CPU resource so I have disabled it and have not
> 
> felt much difference.
> 
> Thanks,
> 
> Kha
> 
>

Re: Bayes filter

Posted by Robert Menschel <Ro...@Menschel.net>.
Hello Khalid,

Tuesday, September 28, 2004, 2:05:51 AM, you wrote:

KW> Dear all,

KW> What is the importance of bayes filter in spamassassin? it takes
KW> a lot of CPU resource so I have disabled it and have not
KW> felt much difference.

IMO Bayes is the most important of the many tools found within
SpamAssassin.  Bayes is the primary reason I can spend my time developing
new rules, since I get only a handful of false negatives a week.

In my view, the main benefits of having an extensive rule set are
1) to help identify new spam that Bayes has not yet learned, and
2) to auto-feed spam to Bayes (since Bayes will not auto-feed itself).

I can understand why Bayes-only anti-spam systems work so well.

Bob Menschel