AutoLearnThreshold

Implements the functionality to submit messages for learning when they fall outside the defined threshold

Example usage

loadplugin      pad.plugins.auto_learn_threshold.AutoLearnThreshold
bayes_auto_learn_threshold_nonspam 0.5 # optional, default is 0.1
bayes_auto_learn_threshold_nonspam 12.0 # optional, default is 12.0
bayes_auto_learn_on_error 1 # optional, default is 1

Usage

When this plugin is loaded after the message has been evaluated by all other plugins it will be evaluated for autolearning. It will be evalute accoring to the following rules:

It calculates the total score for the message from tests that don’t have the noautolearn, userconf tflags

General requirements

  • The autolearn score includes at least 3 body and 3 header tests scores (unless any test has the tflag autolearn_force in which case the header

    and body tests requirement drops to -99)

  • The bayes plugin classified the message differently than this plugin (unless bayes_auto_learn_on_error option is set to 0)

Case 1

  • The message score was higher than the required score
  • The message is considered spam by the autolearn plugin (the autolearn score is higher than the spam threshold)
  • The score from tests with the learn tflag is at least -1

Case 2

  • The message score was lower than the required score
  • The message is considered ham by the autolearn plugin (the autolearn score is lower than the ham threshold)
  • The score from tests with the learn tflag is at least 1

Options

bayes_auto_learn_threshold_nonspam 0.1 (type float)
Messages that score below this value will be submitted for learning as HAM
bayes_auto_learn_threshold_spam 12.0 (type float)
Messages that score over this value will be submitted for learning as SPAM
bayes_auto_learn_on_error 0 (type bool)
Messages will be submitted for learning only if Bayes disagrees with the classification

EVAL rules

This plugin doesn’t expose any eval rules

Tags

None