back to list
Process mining is a technique to extract a process model from an event log. An event log represents a process and consists of traces describing a sequence of activities. Applying process mining techniques to a complex process can lead to a model that is too complex to analyze and reason about. Therefore, analysts are looking at individual variations instead. Traces that follow the same path through the model belong to the same variation. However, the number of variations can be high in complex processes. The analysts’ goal is to analyze a type of behavior. Multiple
variations can describe the same behavior, while only differing slightly. Our solution is to group similar traces into clusters, so they can be investigated as a group. To enable analyses, we visualize the clusters using glyphs on a 2D plane, such that similar clusters are closer to each other. Users
can interact with the visualization by zooming into clusters to analyze, compare, and understand them. Through a number of case studies, we show that our clustering approach succeeds at combining traces that show similar behavior and that our visualization supports users to analyze, compare and understand the clusters.
Project: Visualization of trace clusters
Description
Process mining is a technique to extract a process model from an event log. An event log represents a process and consists of traces describing a sequence of activities. Applying process mining techniques to a complex process can lead to a model that is too complex to analyze and reason about. Therefore, analysts are looking at individual variations instead. Traces that follow the same path through the model belong to the same variation. However, the number of variations can be high in complex processes. The analysts’ goal is to analyze a type of behavior. Multiplevariations can describe the same behavior, while only differing slightly. Our solution is to group similar traces into clusters, so they can be investigated as a group. To enable analyses, we visualize the clusters using glyphs on a 2D plane, such that similar clusters are closer to each other. Users
can interact with the visualization by zooming into clusters to analyze, compare, and understand them. Through a number of case studies, we show that our clustering approach succeeds at combining traces that show similar behavior and that our visualization supports users to analyze, compare and understand the clusters.
Details
- Student
-
RSRik Schreurs
- Supervisor
-
Huub van de Wetering
- Link
- Thesis