hit-frequencies and rules with increased score

[smallrat <to...@gmail.com>]

Hi there,

I wanted to ask, if someone knows how to make hit-frequencies and also
mass-check, to take into account the increased values of rules. 
For examlpe, I have "score bayes_99 (2)" in 50_scores.cf, and I want
./hit-frequencies, and also ./mass-check to deal with this correctly.

How should I do that? 
-- 
View this message in context: http://old.nabble.com/hit-frequencies-and-rules-with-increased--score-tp31575959p31575959.html
Sent from the SpamAssassin - Dev mailing list archive at Nabble.com.

Re: hit-frequencies and rules with increased score

[smallrat <to...@gmail.com>]

John Hardin wrote:
> 
> 
>> I wanted to ask, if someone knows how to make hit-frequencies and also
>> mass-check, to take into account the increased values of rules.
>> For examlpe, I have "score bayes_99 (2)" in 50_scores.cf, and I want
>> ./hit-frequencies, and also ./mass-check to deal with this correctly.
> 
> To what end? The score doesn't determine whether or not the rule hits on a 
> given message. How are the scores on the rules causing hit-frequencies and 
> mass-check to "incorrectly" deal with things? What do you think should 
> happen instead? Please describe what you perceive to be the problem in a 
> little more detail...
> 
> Is it just the display of the rule's score in the output, and this is 
> purely a cosmetic issue?
> 

Hi, thanks for response.

So, here is the situation. I want to create statistics for the rule hits in
my database. I first run mass-check, to create the ham.log and spam.log.

I want to create fp-fn statistic, using ./fp-fn-statistics, but I think,
that these statistics are messed up, if there are not assigned the increased
score values (there are not so many spam hits, as there should be).

As the next step is generating of new scores (only for my three new rules,
the others have "tflags userconf", to not change the scores of these).
So if there are hits, but the corect scores are not assigned (the increased
values), then perceptron won't generate proper scores.

So that's the problem. Actually for set1, set2 and set3, there won't be
proper fp-fn statistics generated, and because I don't change other rule
scores, the fp-fn statistics, that the perceptron generates (according to
the rule scores) won't be correct (the net and bayes rules haven't their
scores from real use assigned, because they are "increased" in our score
files).

-- 
View this message in context: http://old.nabble.com/hit-frequencies-and-rules-with-increased--score-tp31575959p31583112.html
Sent from the SpamAssassin - Dev mailing list archive at Nabble.com.

Re: hit-frequencies and rules with increased score

[John Hardin <jh...@impsec.org>]

On Mon, 9 May 2011, smallrat wrote:

> I wanted to ask, if someone knows how to make hit-frequencies and also
> mass-check, to take into account the increased values of rules.
> For examlpe, I have "score bayes_99 (2)" in 50_scores.cf, and I want
> ./hit-frequencies, and also ./mass-check to deal with this correctly.
>
> How should I do that?

To what end? The score doesn't determine whether or not the rule hits on a 
given message. How are the scores on the rules causing hit-frequencies and 
mass-check to "incorrectly" deal with things? What do you think should 
happen instead? Please describe what you perceive to be the problem in a 
little more detail...

Is it just the display of the rule's score in the output, and this is 
purely a cosmetic issue?

-- 
  John Hardin KA7OHZ                    http://www.impsec.org/~jhardin/
  jhardin@impsec.org    FALaholic #11174     pgpk -a jhardin@impsec.org
  key: 0xB8732E79 -- 2D8C 34F4 6411 F507 136C  AF76 D822 E6E6 B873 2E79
-----------------------------------------------------------------------
   Men by their constitutions are naturally divided in to two parties:
   1. Those who fear and distrust the people and wish to draw all
   powers from them into the hands of the higher classes. 2. Those who
   identify themselves with the people, have confidence in them,
   cherish and consider them as the most honest and safe, although not
   the most wise, depository of the public interests.
                                                   -- Thomas Jefferson
-----------------------------------------------------------------------
  151 days since the first successful private orbital launch (SpaceX)