You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@spamassassin.apache.org by Tom Hendrikx <to...@whyscream.net> on 2014/10/02 11:13:27 UTC
Output of sa-learn --dump magic
Hi,
I am using dspam besides spamassassin, and am interested in comparing
the bayesian data between the two. Dspam reports statistics that include
somewhat standardised metrics for spam filtering: Spam Hit Rate, Ham
Strike Rate and Positive Predictive Value.
I would like to calculate these for spamassassin bayes too, but I need
to know the number of re-learned messages for that, i.e. false positives
and false negatives. The output of sa-learn does not show these, it just
changes the numbers of ham and spam. Are these numbers available to the
spamassassin internals? If so, would it be possible to show these in the
sa-learn output?
Kind regards,
Tom
Re: Output of sa-learn --dump magic
Posted by Tom Hendrikx <to...@whyscream.net>.
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256
On 02-10-14 12:38, Axb wrote:
> On 10/02/2014 11:13 AM, Tom Hendrikx wrote:
>> Hi,
>>
>> I am using dspam besides spamassassin, and am interested in
>> comparing the bayesian data between the two. Dspam reports
>> statistics that include somewhat standardised metrics for spam
>> filtering: Spam Hit Rate, Ham Strike Rate and Positive Predictive
>> Value.
>>
>> I would like to calculate these for spamassassin bayes too, but I
>> need to know the number of re-learned messages for that, i.e.
>> false positives and false negatives. The output of sa-learn does
>> not show these, it just changes the numbers of ham and spam. Are
>> these numbers available to the spamassassin internals? If so,
>> would it be possible to show these in the sa-learn output?
>>
>> Kind regards, Tom
>>
>
> probably not what you want: run sa-learn -D bayes ..... you
> may be able to grep out some info
Not really. I wrote a small wrapper script that writes to syslog when
FPs or FNs are retrained.
Now that I have data, I can give the statistics a go ;)
- --
Tom
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
iQIcBAEBCAAGBQJUMaGoAAoJEJPfMZ19VO/1ht0P/RVSGvUIgH4TVx1v08yGmMq6
cJqOd7QkHCUTcRO0K+7UR3WVI6H/3cVEnftX8oYiYirNh6jKPDpqeklXYIYyFOn8
x0d+CUycYS4FOXiqih5P1Olw9l0Nan0TVjbG60EWyjAfq53BL6I8mvVsuF+aozPU
FQhgK3LgaLwr/6plxaF5KkyFXCqHhfYI14VAdV5klVgZ3ADIgjLTSLBX2v1OIqy4
hFmS1FAfA0RL7N9KVkuW027vvzNjPqyDkrcUgnQWnFs21WUihgDibvD0dBmV5X87
bBJU2uso+ZTdmQtUuAv9v2oBaN37IsUdxY+3m/ZsnEmSNhjy1NGuyolhf3LWvmEo
+QToj1TDtuB3eXIqbPnNoX9rm7qMDooEE0aClo7FZ1Upyu2aXfH+y5jWXd69Oc0b
zZDhZKPc1VVRJXbz1dSSh5BQerIm3K8TNzmPm+1LVLLG4Xf9ItuPyJ6Gm7CMrRWc
0RuQ2ul5p4Q06EEM8nlqCqDe+dPF8oTflTWl+15EjaTDewrNDSpWmaoEuXMAxaUk
8mGOaREkh+UQsCk40ROXHZGF7RilG+iABQIUfP0tGZiHOnfc6VhaqqiwKtwkbqhe
UdWYm8niH7N+6yA3g7uUH8I1awF+D7ZBioFDnn7Bkj1fPM1I+BDVdFdxlW7Bve+l
oVAodF2fc/adKXySZKQW
=T4OC
-----END PGP SIGNATURE-----
Re: Output of sa-learn --dump magic
Posted by RW <rw...@googlemail.com>.
On Thu, 02 Oct 2014 12:38:20 +0200
Axb wrote:
> > these, it just changes the numbers of ham and spam. Are these
> > numbers available to the spamassassin internals?
No
Re: Output of sa-learn --dump magic
Posted by Axb <ax...@gmail.com>.
On 10/02/2014 11:13 AM, Tom Hendrikx wrote:
> Hi,
>
> I am using dspam besides spamassassin, and am interested in comparing
> the bayesian data between the two. Dspam reports statistics that include
> somewhat standardised metrics for spam filtering: Spam Hit Rate, Ham
> Strike Rate and Positive Predictive Value.
>
> I would like to calculate these for spamassassin bayes too, but I need
> to know the number of re-learned messages for that, i.e. false positives
> and false negatives. The output of sa-learn does not show these, it just
> changes the numbers of ham and spam. Are these numbers available to the
> spamassassin internals? If so, would it be possible to show these in the
> sa-learn output?
>
> Kind regards,
> Tom
>
probably not what you want:
run sa-learn -D bayes ..... you may be able to grep out some info