[Spambayes] training problems

Eric S. Johansson esj at harvee.org
Tue Jul 1 16:09:05 CEST 2008


after successfully incorporating spambayes into my project, I'm now having some 
training problems which I suspect are my own fault.  Are there any good 
references on training models?

The particular problem I'm having is that very early on in the training process 
I got the bimodal hump of scores but now I'm getting more and more messages 
scoring in the uncertain zone.  My current training process is everything above 
0.2 is considered good and delivered.  Everything below 0.8 is considered bad 
and dumpstered.  Everything in the middle is presented to the user and what ever 
the user chooses (Spam versus ham) is trained that way.

Like I said, I suspect I'm doing something wrong but I'm not sure what which is 
why I'm looking for training guidance.  Thanks for any education.


---eric



More information about the SpamBayes mailing list