[Spambayes] Important: new feature!

[David Relson]

On Fri, 9 Jul 2004 15:04:40 +1200
Tony Meyer wrote:

> > I understand what you're saying, and I've re-trained 
> > spambayes on equal numbers of HAM and SPAM, 300 each. Yet, it 
> > still has a tendency to classify some HAM as spam. A 
> > significant percentage is like maybe 25% of my regular 
> > announcement mail gets spamflagged...
> 
> Something is definitely not working here - a 25% false positive rate
> is extraordinarily high.  You shouldn't even get as high as 2.5%. 
> Perhaps you could send the spam clues report for one of these messages
> (before any training on it) to the list?  We could take a look at it
> and see if there's an obvious reason that it's failing to be
> classified correctly.

To ask the simple question, are you sure that all your training messages
are correctly classified?  An error or two could have a _big_ effect.