[Spambayes] How low can you go?

Bill Yerazunis wsy at merl.com
Wed Dec 10 11:31:26 EST 2003


   From: Skip Montanaro <skip at pobox.com>

   Okay, time for a little contest.  We've recently seen several users tout the
   size of their training database.  I used to be one of those "enlarged
   database" types, but no more.

   So, how small is yours? <wink>

Well, I'm now running with the mostly-hung CRM114 SBPH/BMM and 
the accuracy is 99.95% or better (most of my errors now are when
a spammer gets onto an email list that has "good credentials"; even
then, if the message is spammy enough, it doesn't get through).

Total size of the training text is 770Kbytes of spam and 570K of
nonspam.  This is something like 250 spams and 150 nonspams, but
that's only approximate.

       -Bill Yerazunis




More information about the Spambayes mailing list