1. Summary
  2. Files
  3. Support
  4. Report Spam
  5. Create account
  6. Log in

bayesian filter - difference to asspanalyze

Moderators: pdreissen, fribo

bayesian filter - difference to asspanalyze

Postby fabiankuelling » Fri Nov 16, 2012 12:50 pm

Hello,

There are mails which are getting full scoring by bayesian filter. When I forward this mails to asspanalyze - the Bayesian Spam Probability is low.

Why that?

Thanks Fabian
fabiankuelling
 
Posts: 28
Joined: Wed May 16, 2012 7:52 am

Re: bayesian filter - difference to asspanalyze

Postby thockar » Thu Nov 22, 2012 8:16 am

This could be happen, if you use the privat spamdb.
Depending on who is forwarding the mail to the analyzer and if (or not) the X-Assp-Intended-For header line is available - the results could be different.

Thomas
thockar
Site Admin
 
Posts: 460
Joined: Mon Mar 09, 2009 7:05 pm

Re: bayesian filter - difference to asspanalyze

Postby fabiankuelling » Fri Nov 23, 2012 8:18 am

Hello Thomas,

What do you mean with "privat spamdb"?

Thanks Fabian
fabiankuelling
 
Posts: 28
Joined: Wed May 16, 2012 7:52 am

Re: bayesian filter - difference to asspanalyze

Postby thockar » Fri Nov 23, 2012 10:42 am

see 'DoPrivatSpamdb'

Thomas
thockar
Site Admin
 
Posts: 460
Joined: Mon Mar 09, 2009 7:05 pm

Re: bayesian filter - difference to asspanalyze

Postby fabiankuelling » Thu Dec 06, 2012 5:08 pm

Hello,

DoPrivatSpamdb is on "NO"

Short extract of my RebuildSpamDB Log - what does the Spam Weight - Not Spam Weight means? Is this to high?

Oct-25-12 10:25:44 Spam Weight: 7,921,841
Oct-25-12 10:25:44 Not-Spam Weight: 7,922,071

Oct-25-12 10:25:44 Corpus norm: 1.0000 - (very good - balanced)
Oct-25-12 10:25:44 Corpus confidence: 1.00000000

Oct-25-12 10:25:44 Total processing time: 4,380 second(s)

Oct-25-12 10:25:44 Total processing data: 1.71 GBbyte

Thanks Fabian
fabiankuelling
 
Posts: 28
Joined: Wed May 16, 2012 7:52 am

Re: bayesian filter - difference to asspanalyze

Postby thockar » Sat Dec 08, 2012 10:30 am

both weights are simple word pair counts for spam and ham

~ 8.000.000 looks (IMHO) too large

how ever if the spamdb and HMMdb are working well, there no reason to worry

Thomas
thockar
Site Admin
 
Posts: 460
Joined: Mon Mar 09, 2009 7:05 pm

Re: bayesian filter - difference to asspanalyze

Postby fabiankuelling » Wed Dec 12, 2012 1:25 pm

Hello,

Can you tell me what the impact of this 8'000 is and how can I reduce it? I remember you had just something around 1'000...

Thanks Fabian
fabiankuelling
 
Posts: 28
Joined: Wed May 16, 2012 7:52 am

Re: bayesian filter - difference to asspanalyze

Postby thockar » Sun Dec 16, 2012 8:27 am

impact: large spamdb and HMMdb

Reduce the number of files in the corpus.

Thomas
thockar
Site Admin
 
Posts: 460
Joined: Mon Mar 09, 2009 7:05 pm

Re: bayesian filter - difference to asspanalyze

Postby fabiankuelling » Fri Jan 11, 2013 10:46 am

Hello Thomas,

I still have this miracle that the bayesian filter tags a mail as SPAM and when I forward the mail (as attachment) to asspanalyze it is ok.

I attached you an example, the mail, the headers of the mail and the asspanalyze... perhaps you can find something....

Could it be that the mysql databases causing this troubles?

Thank you

Fabian
Attachments
Header - Mail.zip
PDF-Files
(125.13 KiB) Downloaded 21 times
fabiankuelling
 
Posts: 28
Joined: Wed May 16, 2012 7:52 am

Re: bayesian filter - difference to asspanalyze

Postby thockar » Sat Jan 12, 2013 1:55 pm

The bayes scanner is the same in both cases!

The results will be different if a rebuild was running between the two checks.

I've done several test - I'm unable to reproduce this.

Thomas
thockar
Site Admin
 
Posts: 460
Joined: Mon Mar 09, 2009 7:05 pm

Next

Return to Troubleshooting

Who is online

Users browsing this forum: No registered users and 1 guest

cron