Naïve Bayesian filters for log file analysis: Despam your logs
Russel W. Havens, Barry M. Lunt, Chia-Chi Teng · 2012
System log files are critical for troubleshooting complex modern computer systems. Systems can easily produce more log file entries than a human can realistically use. However, there are a number of good filtering and clustering technologies that are used in various areas of data mining. This research focuses on using very easily accessible Bayesian spam filters for categorizing log entries. Results of this research have confirmed that these filters can be effectively used to discover log entries related to known issues, and to effectively disprove outage relationships. Both of these techniques can be easily instrumented in a log analysis framework and provide administrators with much needed filtering for similar logs and thus, similar outages.