-
Co-Founder of Vanquish Labs
-
Inventor of Personal Value Control
-
My Bayesian filter is all I need! - Or is it?
Submitted by Frank Urro on Thu, 2007-03-01 16:47.
This author (Ron Lopshire) in usenet group mozilla.support.thunderbird; T-Bird can get you blacklisted, says:
"Let's cut to the chase. Since October, I have received over 21, 000 pieces of spam in my 3 primary non-disposable POP3 accounts. Not one piece of spam got through my Bayesian filter (K9). And that is with only 37 legitimate emails marked as spam (FPs - 0.17%), and most of those should have been marked as spam."
Notice he somehow comes to the logical conclusion that his single phase spam filter is saving him time. Yet, he claims that 37 messages since October (that's 5 messages each month) have been trapped as spam. I read this as - I've sifted through 21,000 pieces of spam (4200 messages each month) looking for trapped legitmate messages.
And that my friends is the crux of the problem - you'll never know what your missing until you look.
For the best anti spam software performance look for a multistage self learning antispam suite.
Far better but still not perfect.
I must admit, since I've made the move to a Bayesian Filter based system (I did my research here at Wikipedia) my acuracies are far better than the other systems I've tried. The one thing I have found it that the Bayesian Filter, given higher acuracies will allow me to interact with my more valuable unsolicited email quicker (because it is not trapped in my spam folder.)
But, you are correct in noting that false positives are still a fact of life so I do feel compelled to check my spam folder now an again. And Bayesian Filter or not, I am occationally reviewing upwardes of 2000 messages on a single day.
I guess it is just a price we must pay for the value of email in general.