[dspam-users] DSPAM training not working - What am I doing wrong ?

[Remi Delon]

Hello,

I've been trying to train DSPAM for a week but to no avail ... Nothing gets 
marked as SPAM.
The way I've been training it is as follows:
- all mails go through procmail, which calls dspam like this:

| dspam --stdout --user=username --deliver=spam,innocent

- when I received a false-negative, I put them in a special folder and then 
I have a script that calls DSPAM on each mail of that special folder like 
this:

| dspam --user username --class=spam --source=error

So I now have the following stats:

dspam_stats -H username
username:
               TP True Positives:              0
               TN True Negatives:           3687
               FP False Positives:             0
               FN False Negatives:           365
               SC Spam Corpusfed:              0
               NC Nonspam Corpusfed:           0
               TL Training Left:               0
               SHR Spam Hit Rate           0.00%
               HSR Ham Strike Rate:        0.00%
               OCA Overall Accuracy:      90.99%

That's what I would expect with my technique (all mails go through DSPAM and 
get marked as TN, unless I later tell DSPAM that it was a FN), but SPAM 
detection still doesn't work (nothing ever gets marked as SPAM and the 
scores are all greater than 0.99 %).

What am I doing worng.

Thanks in advance for any help.

Cheers,

Mike.