bayeslog is a Perl script that filters standard syslog-style UNIX log files using Bayes' Theorem (similar to the way many antispam tools filter out unwanted email messages).
For more information, view the documentation included in the script:
$ perldoc ./bayeslog
Note that bayeslog is very much experimental at this stage.
-
bayeslog
the filter script
-
bayesprint
a debugging tool to print out the current Bayes model from the storage file.
- Install the Algorithm::NaiveBayes module from CPAN.
- Edit the bayeslog script and change the path to the Perl interpreter and the extra library path if required.
- Copy the bayeslog script to a directory in your PATH.
See the integrated documentation or try the --help option.
- FIX: Strip out 'UNSURE XX/YY' from training input lines if present (reentered output).
- Refine word weighting and interpretation.
- Improve selection of log records?
- Change keep/discard terminology to something simpler (good/bad? yes/no?)?
It's experimental, you want the moon on a stick too? :-)