Testing fisher

[Boris 'pi' Piwinger]

Hi!

I have done some extensive testing of fisher constants (two
states). My setting was somehow unusual. First I did a
training with about 15000 hams and 4000 spams. This was
immediately going into production (using -u and manual
corrections). With this live data I was testing my training
database (sic!), which was occasionly enlarged by new ham
and spam mails. So I was not testing bogofilter on new mail,
but on known mail. Here are the results:


algorithm=fisher
min_dev=0.1
spam_cutoff = 0.95

Spam:
   4186 test.spam
False negatives:
364
Ham:
  15140 test.ham
False positives:
1


algorithm=fisher
min_dev=0.2
spam_cutoff = 0.60

Spam:
   4221 test.spam
False negatives:
184
Ham:
  15140 test.ham
False positives:
0


algorithm=fisher
min_dev=0.15
spam_cutoff = 0.60
Spam:
   4237 test.spam
False negatives:
170
Ham:
  15251 test.ham
False positives:
0


algorithm=fisher
min_dev=0.1
spam_cutoff = 0.60

Spam:
   4221 test.spam
False negatives:
139
Ham:
  15140 test.ham
False positives:
1


algorithm=fisher
min_dev=0.075
spam_cutoff = 0.60
Spam:
   4237 test.spam
False negatives:
132
Ham:
  15251 test.ham
False positives:
1

algorithm=fisher
min_dev=0.05
spam_cutoff = 0.60
Spam:
   4237 test.spam
False negatives:
116
Ham:
  15251 test.ham
False positives:
1


algorithm=fisher
min_dev=0.035
spam_cutoff = 0.60
Spam:
   4262 test.spam
False negatives:
101
Ham:
  15251 test.ham
False positives:
1


algorithm=fisher
min_dev=0.025
spam_cutoff = 0.60
Spam:
   4262 test.spam
False negatives:
89
Ham:
  15251 test.ham
False positives:
1


algorithm=fisher
min_dev=0.02
spam_cutoff = 0.60
Spam:
   4297 test.spam
False negatives:
92
Ham:
  15362 test.ham
False positives:
1


algorithm=fisher
min_dev=0.015
spam_cutoff = 0.60
Spam:
   4295 test.spam
False negatives:
92
Ham:
  15361 test.ham
False positives:
1


algorithm=fisher
min_dev=0.0
spam_cutoff = 0.60

Spam:
   4221 test.spam
False negatives:
140
Ham:
  15140 test.ham
False positives:
1


pi