Print Page | Close Window

Speed up bayesian filter learning

Printed From: LogSat Software
Category: Spam Filter ISP
Forum Name: Spam Filter ISP Support
Forum Description: General support for Spam Filter ISP
URL: https://www.logsat.com/spamfilter/forums/forum_posts.asp?TID=4377
Printed Date: 02 June 2025 at 9:00pm


Topic: Speed up bayesian filter learning
Posted By: Guests
Subject: Speed up bayesian filter learning
Date Posted: 01 October 2004 at 11:18am

For our organization, it will likely take a month or more to get up to 5,000 good messages.  Is there a way to expedite the learning of the bayesian filter or plug in someone else's database to get it up and running?

I'm still getting tons of spam that is getting through the keywords and everything else I have set up.  With that in mind, is this effectively training my bayesian filter to recognize these as good messages?


Replies:
Posted By: LogSat
Date Posted: 03 October 2004 at 9:09pm
Steve,

There is an setting in the SpamFilter.ini files to lower the minimum threshold of 5000 god and spam emails to be received. You may try to adjust that setting, however please note that reducing the threshold could reduce the accuracy of the filter.

The spam each organization receives is different from the others, as the customer base is different. A company that has lawyers as customers will receive different kind of emails than one who has doctors as a customer base for example. As such, the statistical corpus database that is created will be different for each case. This said, we've made available for you our own database at logsat.com/spamfilter/pub/temp/corpus.zip. To use, please stop SpamFilter, make a backup of your SpamFilter\corpus directory, then exatract the content of the zip into your corpus directory overwriting your files.

Roberto F. LogSat Software

Posted By: Guests
Date Posted: 04 October 2004 at 11:05am

Thanks, I have installed the already built corpus database from your link.  Unfortunately, I have another problem preventing me from using the SpamFilter now so I won't be able to try it out until I figure out a workaround.  I'll start a new post on that issue to possibly help others in the same situation.

 


Print Page | Close Window