-------- Original-Nachricht --------
> Datum: Thu, 5 Feb 2009 11:28:07 +0100
> Von: "Jehan Pagès" <[email protected]>
> An: [email protected]
> Betreff: [Dspam-user] spam_train does not work

> Hi,
> 
Hallo Jehan,


> I have installed dspam for a postfix/dovecot installation (don't think
> this
> matters, but... if ever), version 3.8.0-r15, installed from the Gentoo
> package manager, built with clamav (not yet activated though), daemon,
> mysql, virtual-users and debug. I am currently trying to train dspam with
> all the spamassassin corpus feed. The command I run (from the man):
> 
>  dspam_train [email protected] spam/ hard_ham/
> 
> I am not in front of the training all the time reading all results, but it
> looks like it always fails on spam and always succed on ham. At least each
> time I was looking at the output of dspam_train, I was seeing such
> behaviour, and this even after having trained thousands of spam/ham. Some
> extract:
> 
> [...]
> [test: nonspam] 00246.fdaacadac7143848978ea0af07 result: PASS
> [test: spam   ] 00491.28cb63173ed4740180e45e6248 result: FAIL (Innocent)
>         [fn] Subject: Finally! Sexy DVDs for FREE. Christmas is Great! NMV
> [test: spam   ] 00492.73db79fb9ad03aff1e08deb73b result: FAIL (Innocent)
>         [fn] Subject: Today's Special: Amazing Penetrations No. 17 29264
> [test: nonspam] 00247.42534d5df0700cb2adf240556c result: PASS
> [test: spam   ] 00493.1c5f59825f7a246187c137614f result: FAIL (Innocent)
>         [fn] Subject: GOV'T GUARANTEED HOME BUSINESS
> [test: spam   ] 00494.fd2efa67e63247ee89cdcf3a6f result: FAIL (Innocent)
>         [fn] Subject: [ILUG] MANUEL OKO
> [...]
> 
> Then if I look the stats of the user I am training:
> 
> # dspam_stats -H [email protected]
> [email protected]:
>                 TP True Positives:              0
>                 TN True Negatives:           5619
>                 FP False Positives:             0
>                 FN False Negatives:          4745
>                 SC Spam Corpusfed:          14920
>                 NC Nonspam Corpusfed:           0
>                 TL Training Left:               0
>                 SHR Spam Hit Rate           0.00%
>                 HSR Ham Strike Rate:        0.00%
>                 OCA Overall Accuracy:      54.22%
> 
> First we not that it fails around half the time here, but also that the
> nonspam corpusfeed is null! As though I never fed the trainer with any ham
> (even though I did, as you can see in the above extract). I tried to
> search
> on the web for such issue and found some similar problems, but nothing was
> clearly given (or I searched with wrong terms). Could you help me to
> understand this out and make my training succeed? Currently in my test
> mailbox, all received spams are always considered "innocent" (in fact I
> still never had any email flagged spam, even after all this training). Is
> it
> an issue of configuration?
>
Who knows? You did not send the configuration you are using. How about doing 
that?


> A bug?
>
Might be. But more information is needed in order to be sure.


> Thanks all.
> 
> Jehan
>
// Steve
-- 
Psssst! Schon vom neuen GMX MultiMessenger gehört? Der kann`s mit allen: 
http://www.gmx.net/de/go/multimessenger01

------------------------------------------------------------------------------
Create and Deploy Rich Internet Apps outside the browser with Adobe(R)AIR(TM)
software. With Adobe AIR, Ajax developers can use existing skills and code to
build responsive, highly engaging applications that combine the power of local
resources and data with the reach of the web. Download the Adobe AIR SDK and
Ajax docs to start building applications today-http://p.sf.net/sfu/adobe-com
_______________________________________________
Dspam-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/dspam-user

Reply via email to