Change Detection in Event Logs by Clustering

摘要

The detection of changes in event logs recording the behavior of flexible processes is especially challenging and process mining algorithms generate useless "spaghetti" models out of them. Due to this, existing approaches for change detection in event logs recording the behavior of flexible processes can only localize a change point, which is of no avail when it comes to explain when, why and how a process model changed and will change. The aim of this paper is to present a novel clustering technique laying the foundation to determine a variety of changes and to foresee changes. In order to do this, four algorithms have been developed. We report the results of evaluations on synthetic as well as real-life data demonstrating the efficiency of the approach and also its broad scope of application for event and sensor data.