All the Perl that's Practical to Extract and Report

Bayesian

davorg on 2004-04-25T10:24:23

I was wondering why the amount of spam in my inbox had increased drastically over the last month or so. Further investigation shows that for some reason the bayesian filters had stopped working.

So yesterday I spent some time sorting all that out. Overnight I retrained it from large piles of spam and ham and now everything seems to work fine again. It's good to see entries in my spam folder saying things like

 pts rule name              description
---- ---------------------- --------------------------------------------------
 5.4 BAYES_99               BODY: Bayesian spam probability is 99 to 100%
                            [score: 1.0000]

Bayesian flooding

dws on 2004-04-25T16:40:06

Interesting. I've also noticed a large number of false negatives, and have been wondering whether keyword flooding has been spoiling the data. After you retrained, have your non-spams been scoring differently?