Spam Statistics

Spam Statistics

Notes:

• in early December 2004 I joined the high-volume Fedora Users mail-list (averages about 400 messages a day, most of them crap). Thus the message counts start to skyrocket, and the proportion and number of non-spam messages also skyrockets.
• on 12 December 2004, the process that updates the bayesian learning blew out its database and now has to be retrained. So the 'Escaped Spam' field will only be filled in from that point forward. Perhaps I should reset the statistics so we can watch the learner in action.

• I reset the logs, removed my corpus, and deleted the one-day-old bayesian dictionaries.
• I'll keep the spam and ham for a while, and we can see how things progress.
• For now let's also say I'll train the bayesian filter and update the virus scanner weekly.
• I'll also set up a cronjob to capture statistics every monday or so and we can see how fast the bayesian filter learns.