[Spambayes] runratio with timcv.py
Brad Clements
bkc@murkworks.com
Wed, 09 Oct 2002 16:09:50 -0400
Hmm,
Well I didn't get tim's message about "using timcv.py for incremental is bad" until after I
started my ratio testing.
This is timcv.py on -n 10 with 1200 messages total per set.
use_central_limit: True
I also have timtest.py running, I have no idea if runratio.sh will handle it's output.. Will
post when that finishes running.
Also, I modified runratio.sh to handle arbitrary list of spam/ham count steps.. want me
to post?
(last stat line)
-> <stat> tested 1050 hams & 150 spams against 9450 hams & 1350 spams
And the table
ham-spam: 150-1050 300-900 450-750 600-600 750-450 900-3001050-150
fp tot: 30 39 45 48 48 50 40
fp %: 2.00 1.30 1.00 0.80 0.64 0.56 0.38
fn tot: 14 20 17 19 14 16 15
fn %: 0.13 0.22 0.23 0.32 0.31 0.53 1.00
h mean: 3.31 2.36 1.93 1.74 1.53 1.36 1.08
h sdev: 13.41 10.95 9.86 9.47 8.86 8.26 7.31
s mean: 99.37 99.16 99.02 98.74 98.57 98.26 97.11
s sdev: 5.61 6.50 7.04 8.08 8.54 9.46 12.29
mean diff: 96.06 96.80 97.09 97.00 97.04 96.90 96.03
k: 5.05 5.55 5.74 5.53 5.58 5.47 4.90
Brad Clements, bkc@murkworks.com (315)268-1000
http://www.murkworks.com (315)268-9812 Fax
AOL-IM: BKClements