[Spambayes] More spam than ham

[Gary Smith]

Thank you for your reply.

I still have a fair amount of spam going to unsure, .ome 
going to ham and some ham listed as unsure.

I have set my Pegasus filters to move unsure into the 
unsure folder. By doing this, I see a very minimum of 
spam get through to my "new Mail" folder. I train on 
the unsure to classify spam as spam yet sometimes 
ham gets wrongly assigned as unsure or spam. 

I look forward to those rare misclassifications of ham 
because they mitigate the ratio of unsure spam & ham 
listed as spam (which are always retrained).

I suppose what I need to know is if there's an 
imbalance between the spam/ham ratio, does that 
affect the accuracy of Spambayes?

After all is said & done, more spam than ham needs to 
be corrected and the ratio seems destined to be 
skewered to the left.

Gary

> If you're confident about the effectiveness of spambayes, then the best
> thing to do is disable training and use the current settings. Of course
> you definitely need to double check your unsure and spam folders to
> recover the ocasional false positive but they should be few & far
> between at this point.
> 
> On 19 May 2006 at 13:48, Gary Smith wrote:
> 
> > A twist for me,
> > 
> > Spambayes has been so good at finding HAM that it no 
> > longer identifies HAM as unsure or SPAM. I'm still 
> > getting spam that sometimes shows as HAM so I train 
> > it as such.
> > 
> > Now I get this notice:
> > 
> > "Warning: you have much more spam than ham - 
> > SpamBayes works best with approximately even 
> > numbers of ham and spam."
> > 
> > How do I deal with this? should I find old emails known 
> > to be good and paste the raw header/message info 
> > into the "Train on a message" window?
> > 
> > Since so much is accurately coming in as HAM but 
> > SPAM is eternaly new, how do I best bolster the 
> > trained HAM emails ratio?
> > 
> > Thanks,
> > 
> > Gary
> > _______________________________________________
> > SpamBayes at python.org
> > http://mail.python.org/mailman/listinfo/spambayes
> > Check the FAQ before asking: http://spambayes.sf.net/faq.html
> 
> 
> 
> -----BEGIN PGP SIGNATURE-----
> Version: PGP 8.0.3 - not licensed for commercial use: www.pgp.com
> 
> iQA/AwUBRG9Kia5LcfzXROjOEQJyOACfbZVDlmzrO1I8GCQ/ATjJ/k5j1yUAoP1r
> jq+osO6MG6lg4o70+rI4VGNS
> =hdVz
> -----END PGP SIGNATURE-----
> 
>