[Spambayes] False spam identification

[Skip Montanaro]

On Tue, Mar 18, 2014 at 9:12 PM, Fred Smith
<fredex at fcshome.stoneham.ma.us> wrote:
> Once you train it on enough of them ...

I will also point out one potential issue which we used to see BITD.
Don't train SpamBayes on every single message, only on the ones that
aren't classified correctly. You wind up with a much smaller training
database, are less likely to misclassify messages, and will likely see
stronger effects on the classification accuracy in the situations
where you do add a new message.

Skip