Hi, 1) It's good to have more ham than spam because false positives are more anoying than false negatives. 2) It's good to keep both spam and ham "fresh" since they both tend to change over time.
Cheers, Daniel. On Tue, Jul 29, 2003 at 03:31:14PM -0500, Thomas Cameron wrote: > All - > > I have trained Bayes using about 1000 ham and 2000 spam. As time goes on, I > am feeding Bayes false negatives (spam that slipped through). If I get to > the point where there are several thousand spam and only the original 1000 > ham will it screw up Bayes? In other words, do I need to keep the ham:spam > ratio even or anything like that? > -- > Thomas Cameron, RHCE, CNE, MCSE, MCT > Cameron Technical Services, Inc. > http://www.camerontech.com > (512) 454-3200 > > > > ------------------------------------------------------- > This SF.Net email sponsored by: Free pre-built ASP.NET sites including > Data Reports, E-commerce, Portals, and Forums are available now. > Download today and enter to win an XBOX or Visual Studio .NET. > http://aspnet.click-url.com/go/psa00100003ave/direct;at.aspnet_072303_01/01 > _______________________________________________ > Spamassassin-talk mailing list > [EMAIL PROTECTED] > https://lists.sourceforge.net/lists/listinfo/spamassassin-talk -- Daniel Carrera | PGP: 6643 8C8B 3522 66CB D16C D779 2FDD 7DAC 9AF7 7A88 Math PhD. UMD | http://www.math.umd.edu/~dcarrera/pgp.html A child's letter to GOD: Dear GOD, Did you really mean "do unto others as they do unto you?" Because if you did, then I'm going to fix my brother. -- Darla ------------------------------------------------------- This SF.Net email sponsored by: Free pre-built ASP.NET sites including Data Reports, E-commerce, Portals, and Forums are available now. Download today and enter to win an XBOX or Visual Studio .NET. http://aspnet.click-url.com/go/psa00100003ave/direct;at.aspnet_072303_01/01 _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
