Well, I can give you any fixed ratios, but I can tell you that this "score deviation" is likely to change when the GA is run. It's likely to change very substantially if the GA itself changes, as happened between 2.31 and 2.4x, and it's likely to deviate more mildly as the content of the corpus and ruleset changes over time.

Really your best idea of what the spread for a given GA run is like is going to come from reading the top of the STATISTICS.txt. This will show things like average spam score, average nonspam score, average FP, average FN, etc. Comparing that information between releases will likely give you the best "feel" for how much to adjust your thresholds.

At 12:44 PM 10/23/2002 -0400, [EMAIL PROTECTED] wrote:
Hello,

I'm trying to determine how many spamassassin hits are appropriate for
tagging messages at different levels of it possibly being spam.

I have spamassassin 2.31 running on a production machine, and figured out
some good numbers for that situation. I installed 2.43 on a different
machine, and it's producing MUCH different results.

With the sample-spam.txt that ships with 2.43:
        2.31 returns 30.1 hits
        2.43 returns 14.7 hits
These were both run in the same way (spamassassin -t < sample-spam.txt)

The results output shows much lower point definitions for the items it
finds in 2.43, but both match the same 16 rules.

Can anyone offer some advice on how much existing filters for hit numbers
should be adjusted? For example, a score of 7 on 2.31 corrolates to X.X on
2.43? ...if I devide all my current client hit settings by 2.047 (ratio from
30.1/14.7) will I get the same results I'd expect from the previous version?

I was also wondering if the hit numbers returned normally change this much
much between releases. Is this something that will need adjusted on every
release? or was this something that changed once and isn't likely to
change for a while?

Thanks for your advice,
--
Josh I.
(if this is answered in a FAQ somewhere, please let me know, I couldn't
find it)



-------------------------------------------------------
This sf.net email is sponsored by: Influence the future
of Java(TM) technology. Join the Java Community
Process(SM) (JCP(SM)) program now.
http://ads.sourceforge.net/cgi-bin/redirect.pl?sunm0002en

_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk


-------------------------------------------------------
This sf.net email is sponsored by: Influence the future of Java(TM) technology. Join the Java Community Process(SM) (JCP(SM)) program now. http://ads.sourceforge.net/cgi-bin/redirect.pl?sunm0002en

_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk

Reply via email to