[Spambayes] SpamBayes compared to POPFile

[Skip Montanaro]

    David> Once trained, POPFile seems to do a pretty good job. I haven't
    David> actually gathered any statistics to tell me which of the two is
    David> the most accurate but my feeling is that SpamBayes just edges
    David> it. More importantly for me, POPFile produced a number of false
    David> positives (ham wrongly put into the spam category) which I find
    David> unacceptable - since I got SpamBayes I have got used to skimming
    David> over my Spam folder much more quickly and less often, as I can
    David> rely on there being nothing important in there. I think it's
    David> probably the concept of an "unsure" category in SpamBayes that
    David> gives it the edge here.

Perhaps a good way to run would be to use SB to distiguish ham from spam
(and unsure), then only feed your putative ham into POPFile.

Skip