Barca: Bayesian and General settings in Junk Mail

[saoir]

There are loads of discussions spread all over the place about all aspects of the filtering. But too many are based on individuals specific problems.

If I may - I would like to start a thread based specifically and solely on the general meaning, and best values of, this set of settings.

My settings are:

Custom Sensitivity: 10 (this appears to be the threshold above which a score defines the mail as Junk and is moved to the Junk Mailbox ?)

Junk Threshold: 0.99
Good Mail Bias: 1.00
Junk Score: 20.00
Good Score: 1.00

First question: What do all of these setting mean ?

Second question: Is it possible to reduce the sensetivity level below 10 ? Mail that used to be caught by my filters is now not being filtered and most of it has a value of 8 or 9. I tried to change the number in the .ini file but it always changes back when I go back to Barca

[SFCurley]

If you try to change the ini file while poco is open, it will always get overwritten b/c the settings in memory will be saved over. If you want to change the sensitivity in the ini file directly, you'd have to close PM first. Then the settings will stick. You can also change w/ Poco open by adjusting the threshhold slider bar by doing Ctl-F4, then General Settings. Should stick that way, too.

[saoir]

You can also change w/ Poco open by adjusting the threshhold slider bar by doing Ctl-F4, then General Settings. Should stick that way, too.

But it won't go below 10 that way.

[FieldDir121]

One of the things I noticed with the Bayesian filter was that too much of the header was added to the DBgood.ini and DBspam.ini file contents. I never tried it, but it seemed the spam scoring would be more accurate if the header was just ignored with the possible exception of the subject. Too much of all headers is boiler plate.

Here are some examples from DBgood.ini:
X-POCOSYSTEMS-MAILSCANNER-clean=2
X-MSMAIL-PRIORITY-x-msmail-priority=10
X-MIMEOLE-v6.00.2900.2180=2
USER-AGENT-1.0=4

And from DBspam.ini:
X-POCOSYSTEMS-MAILSCANNER-clean=1
X-MSMAIL-PRIORITY-x-msmail-priority=6
X-MIMEOLE-v6.00.3790.181=1
USER-AGENT-1.0=2

And why should this be in either file?
DATE-13=14
CONTENT-TYPE-related=2
$12=1
Judging spam based on:
The date?
Text versus html?
The price of the item being offered?

These are what have lead me to abandon the Bayesian filter and attempt to create my own. Managing my own filters takes time. Time I used to spend looking through the messages the Bayesian filter decided were spam when not all of them were.

To each his own.

Scott

[SFCurley]

I think that all of the header info that comes in on an email should be used for Bayesian purposes, but header elements that Poco adds should not be. Popfile, which I think is pretty close to a gold standard in terms of effectiveness, does use most, if not all, header data, along with other things called pseudotags. That said, however, it may not matter much, if at all, whether or not Poco looks at elements like FieldDir121 illustrated because only the top 30 most indicative spam or good tokens are used in Poco's Bayesian calcsulations and it's quite likely that those shown would not make the top 30 list.

[Guest]

The following is how I solved this problem.

If you can't lower the POCO value, raise the number for a positive.

CSSJR

Date: Sun, 12 Mar 2006 02:05:50 -0600
Delivery-Date: Sat, 11 Mar 2006 19:31:15
Status: U
X-Poco-Spam-DNSBL: Received from IP address (203.215.249.24)
sbl-xbl.spamhaus.org-true (+25)
spamsources.fabel.dk-false (+0)
bl.spamcop.net-true (+25)
list.dsbl.org-true (+25)
dnsbl.njabl.org-true (+25)
dnsbl.sorbs.net-true (+25)
Debug - v1.19 #spamscore 125 #waitattempt 7 #timeout 10 #mode 1
X-Poco-Scored: +125
X-Poco-Score-Detail: -5 [%BAYES%=P=0;T=90;BIAS=+20] (%bayes% P=0;T=90;Bias=+20)
X-Poco-Score-Detail: +3 [X-MAILER=] (X-Mailer )
X-Poco-Score: +125
X-Poco-Score-Detail: +2 [FROM=%ADDRESSBOOKS%] (From %addressbooks%)
X-Poco-Scored: +125
X-Poco-Score-Exceeds: 10
Subject: All Investors, L@@K For Big Returns! r33
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
X-Poco-UID: 23802299
X-Poco-Status: U