Konferensartikel

We present our work on a visual, similarity-based approach to log file exploration. The use of similarity rather than simple aggregation schemes empowers users to focus on the high-level events behind log entries, rather than the entries themselves. We make use of an accelerated version of TRIAGE to determine the similarity coefficients for each pair of log entries. The model is embedded in an interactive visualization system which enables the fluid interpretation of similarities with the help of a simple clustering approach.

Clustering; similarity measures