On Wednesday, September 15, 2004, 3:26:07 PM, Nick wrote: NH> _M:
>> > I recommend that you review your weights using a combination of >> > your current experiences and the spam test quality analysis found >> > here: >> > >> > <http://www2.spamchk.com/public.html> NH> Is the code to make these graphs avail? I saw a thread before about NH> this before but do not recall the outcome.. They are really nice.. Not quite, but I'm working on it. What I do have is MDLP V0.9 which is in testing. This utility will produce the source data for these graphs from a Declude MID log file. The output is xml, or a family of csv files. If you would like to be on the testing team for this utility please let me know at [EMAIL PROTECTED] I think I might build a function for generating these graphs from the MDLP utility in the first version - but they won't be quite as nice. (MDLP = Modular Declude Log Processor) On or about V2.0 of this utility it will be able to automatically tune and maintain Declude weights based on this analysis & some AI I'm building. >> > One formula that you can use to derive your test weights from this >> > analysis is W = (SA^2)*HOLD_WEIGHT. So, in the case of these two >> > groups you might select these weights for your system: NH> What are your estimated accuracies for the various rule groups? I think the numbers on the site above are pretty good. Doing the math is pretty straight forward - simply square the SA (or HA for a ham test) to get the magnitude of the weight you should use. Then you might nudge these numbers one direction or another depending upon how you feel about them. Hope this helps, _M This E-Mail came from the Message Sniffer mailing list. For information and (un)subscription instructions go to http://www.sortmonster.com/MessageSniffer/Help/Help.html
