Ham to Spam ratio
Tom Anderson
tanderso at oac-design.com
Wed Jan 26 16:33:39 CET 2005
----- Original Message -----
From: "Johannes Klug" <derjoi at gmx.net>
To: <bogofilter at bogofilter.org>
Sent: Wednesday, January 26, 2005 10:28 AM
Subject: Ham to Spam ratio
> Hello!
>
> I retrain bogofilter from my collected ham and spam corpus every few
> months,
> to build up a fresh database.
> My ham collection now is about 1k, spam about 3k.
> Does this ratio have an impact on bogofilter's performance? Should I train
> with 1k both?
It doesn't seem to effect mine and I've been seeing at least 10:1 spam to
ham ratio for like two years. Then again, I do exhaustive training (until a
message scores appropriately), which may counteract any natural affect of a
skew.
Tom
More information about the Bogofilter
mailing list