Visualization of Server Log Data for Detecting Abnormal Behaviour

Modern large-scale IT operations are getting complex every day and generate thousands of log and event files, which contain essential information about the systems and their behavior. System errors are examined manually by human experts, which takes a considerable amount of time and efforts. The system log files, besides other attributes such as time and location, contain messages in textual form, which is essential for analyzing logs behavior. In this paper, we propose a method to extract meaningful information through topic modeling from log messages, combine them into a sequence of events and present them to the practitioners to visually identify abnormal behavioral patterns, which is used for predictive analysis.