>-----Original Message----- >From: Max Kipness [mailto:[EMAIL PROTECTED] >Sent: Thursday, May 20, 2004 2:15 PM >To: [EMAIL PROTECTED] >Subject: RE: Feedins spam with no ham? > > >>I've got about 1000 spam messages saved up, and was wondering if it >>would have an adverse affect on Bayes if I fed these to it, without >>feeding any ham? >> >>I currently have autolearn on, and it's doing pretty good, >but I'd like > >>to get it better. >> >>Thanks, >>Max >> >> > >>It won't hurt it much, but 50/50 should be the optimum goal. > >>--Chris > > >Thanks for the reply. > >But am I better off not feeding it that spam and waiting until I can >collect an equal amount of ham?
No, you want to feed it!!! Feed it until you here it burp. > >Right now, as I said, it's doing a pretty good job with autolearn. You >say it won't hurt much, do you mean it may increase undetected spam? > It will definetly help. I just like to show that it is a statistical equation. They can be skewed with data that is skewed. But in the case of spam, it needs to know what it missed. Or what is prevelent. But do try to give it a propper diet of ham. --Chris
