#926 statistics: spam only

[mnent]

I have the following suggestion:

From all statistical analysis done within POPfile, the
accuracy with regard to the spam folder is the most
important to me. This is because "false positives" for
this bucket are the most dangerous. If POPfile doesn't
get it right for one of the other buckets, this is less
important to me because I will look into these mail
folders anyway.

However, if I had a separate figure of the accuracy
just for spam that tells me "99,80%" (which I think is
quite likely), I would be inclined to consider not to
check the spam folder any more "just to be on the safe
side".

When I first used POPfile about a year ago, I had only
two buckets (spam and OK) and accuracy soon rocketed to
something like 99,20%. But since I have more than these
two buckets (in fact only a third one for all
list-related stuff), accuracy dropped to around 97%. I
can live with this (in particular, as this is much much
better than any ratio I hear from colleagues who don't
use POPfile but other antispam software!). However, as
I said, filtering through a hundred spam mails per day
is tyring and I would like to get a good argument in
favour of not doing it anymore.

So my suggestion is:
Don't give only one overall percentage for total
accuracy, but also a second, separate percentage for
just the false positives in the spam bucket in relation
to the rest of all messages. (Note that even the "false
negatives" in regard to spam should not be included in
this figure as this is no problem , whereas missing an
important message in the spam folder is).

Thanks for considering this!
Michael