Hello Roelof, Wednesday, September 29, 2004, 11:31:59 AM, you wrote:
> Furthermore as I've only recently started to use BayesIt, I've got > this question. I understand that I've got to mark whatever I consider > unwanted as Junk, that's the easy part. While BayesIt is learning it > lets everything pass, Not really. It filters to the best of it's ability. It will make many mistakes due to unsufficiant data but tends to create much more false negatives than false positives. > does this mean that I've got to mark everything > that's okay (and not marked as junk by BayesIt) explicitly as 'not > junk' or is that done by the plug-in since I don't mark it as junk? Train only on mistakes. Mark as junk spam that doesn't make it into the junk folder. Mark as not junk ham that appears in the junk folder. > During the 48 hours that I got BayesIt installed, I got 7 spams and > 940 good mails, is that a ratio the plug-in can improve upon? You'll need much more spam. I'd recommend at least 500 spams to train it. That's why it's good to keep a certain amount of spam mails if the neccessity appears to retrain bayesit or train a new bayesian filter. -- Best regards, Andre ________________________________________________________ Current beta is 3.01 RC/1 | 'Using TBBETA' information: http://www.silverstones.com/thebat/TBUDLInfo.html IMPORTANT: To register as a Beta tester, use this link first - http://www.ritlabs.com/en/partners/testers/
