bogotune results

[Greg Louis]

> >>Recommendations:
> >>
> >>---cut---
> >>db_cachesize=4
> >>robx=0.600000
> >>min_dev=0.020
> >>robs=0.0100
> >>spam_cutoff=0.069       # for 0.05% fpos (1); expect 0.00% fneg (0).
> >>#spam_cutoff=0.040      # for 0.10% fpos (2); expect 0.00% fneg (0).
> >>#spam_cutoff=0.020      # for 0.20% fpos (4); expect 0.00% fneg (0).
> >>ham_cutoff=0.020
> >>---cut---
> >
> >
> >So why does bogotune suggest those values? Zero false
> >positives were easily possible with a higher ham_cutoff.
> >
Bogotune can't look for zero false positives; it has to look for at
least one.  If you have 10,000 messages or more, it will try to find a
0.01% value.  If not, it does the best it can with what you give it. 
Obviously one can (and I usually do) just use the recommendation as a
starting point from which to set a comfortable balance between fp and
fn.

You can't influence false positives with the ham cutoff; it sets the
boundary between nonspam and unsure.  A false positive is a nonspam
that scores above the spam cutoff.

-- 
| G r e g  L o u i s         | gpg public key: 0x400B1AA86D9E3E64 |
|  http://www.bgl.nu/~glouis |   (on my website or any keyserver) |
|  http://wecanstopspam.org in signatures helps fight junk email. |