[Spambayes] CL2 part I (reprise)

Brad Clements bkc@murkworks.com
Sun, 06 Oct 2002 13:51:17 -0400


FP rate is down after cleaning up training sets

but CL2 seems less sure than CL3

-> <stat> Ham scores for all runs: 6500 items; mean 1.36; sdev 8.88
-> <stat> min 0; median 0; max 100
* = 104 items
  0 6339 *************************************************************
 25  100 *
 50   44 *
 75   17 *

-> <stat> Spam scores for all runs: 6500 items; mean 99.27; sdev 6.21
-> <stat> min 0; median 100; max 100
* = 106 items
  0    4 *
 25   14 *
 50   74 *
 75 6408 *************************************************************
-> best cutoff for all runs: 0.5
->     with weighted total 1*61 fp + 18 fn = 79
->     fp rate 0.938%  fn rate 0.277%

Reading results/cl2-a/clim.pik ...
Nham= 6500
RmsZham= 4.91231538116
Nspam= 6500
RmsZspam= 3.7454876964
======================================================================
HAM:
FALSE POSITIVE: zham=-9.71 zspam=-2.38 Data/Ham/Set1/10180 SURE!
FALSE POSITIVE: zham=-6.08 zspam=-2.50 Data/Ham/Set1/10852 SURE!
FALSE POSITIVE: zham=-6.09 zspam=-2.19 Data/Ham/Set3/6480 SURE!
FALSE POSITIVE: zham=-4.08 zspam=-1.36 Data/Ham/Set4/69 SURE!
FALSE POSITIVE: zham=-4.32 zspam=-2.28 Data/Ham/Set4/5548 SURE!
FALSE POSITIVE: zham=-10.65 zspam=-1.44 Data/Ham/Set4/10008 SURE!
FALSE POSITIVE: zham=-5.43 zspam=-2.03 Data/Ham/Set4/10937 SURE!
FALSE POSITIVE: zham=-10.25 zspam=-0.34 Data/Ham/Set5/5105 SURE!
FALSE POSITIVE: zham=-4.61 zspam=-2.40 Data/Ham/Set5/6369 SURE!
Sure/ok       6280
Unsure/ok     184
Unsure/not ok 27
Sure/not ok   9
Unsure rate = 3.25%
Sure fp rate = 0.14%; Unsure fp rate = 12.80%
======================================================================
SPAM:
FALSE NEGATIVE: zham=-1.00 zspam=-5.48 Data/Spam/Set2/5185 SURE!
FALSE NEGATIVE: zham=-2.01 zspam=-4.62 Data/Spam/Set2/6457 SURE!
FALSE NEGATIVE: zham=-2.09 zspam=-18.28 Data/Spam/Set3/3010 SURE!
FALSE NEGATIVE: zham=-2.46 zspam=-4.49 Data/Spam/Set4/6367 SURE!
FALSE NEGATIVE: zham=-2.46 zspam=-4.49 Data/Spam/Set4/6371 SURE!
Sure/ok       6226
Unsure/ok     224
Unsure/not ok 45
Sure/not ok   5
Unsure rate = 4.14%
Sure fn rate = 0.08%; Unsure fn rate = 16.73%



Brad Clements,                bkc@murkworks.com   (315)268-1000
http://www.murkworks.com                          (315)268-9812 Fax
AOL-IM: BKClements