Hi all,
 Bayes seems to be missing quite  a lot of spam. I'm getting these results quite often:
 
TOP SPAM RULES FIRED
----------------------------------------------------------------------
RANK    RULE NAME                       COUNT  %OFMAIL %OFSPAM  %OFHAM
----------------------------------------------------------------------
   1    HTML_MESSAGE                     2127    68.95   75.29   62.00
   2    URIBL_BLACK                      1822    39.09   64.50   11.22
   3    URIBL_SC_SURBL                   1664    31.07   58.90    0.54
   4    URIBL_OB_SURBL                   1654    31.61   58.55    2.06
   5    BAYES_00                         1471    65.28   52.07   79.77
   6    URIBL_SBL                        1360    29.81   48.14    9.70
   7    URIBL_WS_SURBL                    922    17.37   32.64    0.62
   8    AWL                               911    42.81   32.25   54.39
   9    URIBL_AB_SURBL                    746    13.89   26.41    0.16
  10    BAYES_99                          707    13.09   25.03    0.00
 
To me, it looks like Bayes_00 is hitting far too much spam.
 
I have fed a large amount of mail into Bayes:
 
[EMAIL PROTECTED] ~]# sa-learn --dump magic
0.000          0          3          0  non-token data: bayes db version
0.000          0       6468          0  non-token data: nspam
0.000          0       6471          0  non-token data: nham
0.000          0     160969          0  non-token data: ntokens
0.000          0 1150774613          0  non-token data: oldest atime
0.000          0 1153439019          0  non-token data: newest atime
0.000          0 1153436831          0  non-token data: last journal sync atime
0.000          0 1153426735          0  non-token data: last expiry atime
0.000          0    1382400          0  non-token data: last expire atime delta
0.000          0      97882          0  non-token data: last expire reduction count
And I'm quite certain that it was fed correctly.
All of the misses I have checked have hit Bayes_00.
 
Any ideas why this is happening? I have toyed with the idea of lowering the bayes_00 score. Anyone care to enlighten me on whether this would be a bad idea and why?
 
 
Regards,
             Leigh
 
Leigh Sharpe
Network Systems Engineer
Pacific Wireless
Ph +61 3 9584 8966
Mob 0408 009 502
email [EMAIL PROTECTED]
web www.pacificwireless.com.au
 

Reply via email to