On Sunday, December 28, 2003, at 01:40 PM, Simon Byrnand wrote:
I upgraded from 2.60 to 2.61 and I am getting many false positives. It seems that Bayes is pushing it with a score of 5.4. What are people to do to get around this? Do you set Bayes for a lower score? Do you disable? Thanks!
5.4 BAYES_99 BODY: Bayesian spam probability is 99 to 100% [score: 1.0000]
If you're getting BAYES_99 on a FP (ham) then something is wrong with your training of the bayes database. Do you just use autolearn and/or do you manually train it ?
That's a good question. I used whatever was the default for installing SA. I'm still a newbie to SA. Where do I look to find the answer to your question?
Well, if you don't know, then you're probably just using autolearn - which is on by default. To manually learn messages as ham or spam you use the sa-learn command, so if you're not using that, you're not doing any manual learning.
I also found out BAYES_00 scoring another false positive with 4.9.
Err, BAYES_00 is *minus* 4.9 :) And is correct if the message is ham. If it still got flagged as spam, it must have had many other tests that fired...without seeing the headers, we can only speculate...
I just assumed BAYES was all about spam, it is clear from what you pointed out I need to read more about what Bayes does. Thanks!
Bayes (in simple terms) works by looking at and comparing the word statistics between spam and ham - if it doesn't learn ham as well it has nothing to go by. Generally its a good idea to learn roughly equal amounts of spam and ham, although since version 2.60 its reasonably tolerant of quite a wide imbalance....
Lower numbered bayes scores between BAYES_00 and BAYES_30 actually mean it thinks its ham, and give negative scores, while BAYES_60 to BAYES_99 mean it thinks its spam, and give positive scores. In between is uncertain.
If a ham or spam is showing up with the opposite BAYES indication to what it should be, it's a good idea to learn it manually. (How you go about doing that is going to depend on how and where you read your email.....)
Regards, Simon
------------------------------------------------------- This SF.net email is sponsored by: IBM Linux Tutorials. Become an expert in LINUX or just sharpen your skills. Sign up for IBM's Free Linux Tutorials. Learn everything from the bash shell to sys admin. Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
