[bogofilter-announce] bogofilter-0.13.5 - new current release
Boris 'pi' Piwinger
3.14 at logic.univie.ac.at
Wed Jun 4 13:45:25 CEST 2003
Greg Louis wrote:
>> robs=0.001
>> 111
>> False positives:
>> 4
>> ------------------------------------------
>> robs=0.01
>> False negatives:
>> 90
>> False positives:
>> 5
>> ------------------------------------------
>>
>> This again shows for my messages, that the new robs is worse.
>
> It shows nothing of the sort. It shows that you didn't adjust the spam
> cutoff, which must be done, because changing s changes the distribution
> of scores. That's all it shows.
>
> When comparing different s values, you need to adjust the spam cutoff
> so your fp rate stays the same, then compare fn. Otherwise the
> comparison is meaningless.
Well, as I said, all the false positives have values of >.85
for the former and >.9 for the latter. Adjusting the cutoff
to cope with that would greatly increase false negatives.
pi
More information about the Bogofilter
mailing list