Visually Representing History Dependencies in Event Logs

Many process mining tools produce directly-follows graphs (DFG) as visual representations of event logs. While the “directly follows” relation is a good starting point for visualizations, there are simple phenomena it does not capture, for instance, when whether or not an event directly follows another event depends on the event directly preceding it. We call this a history dependency. This paper presents an empirical study of preferences for visualizing history dependencies: plain DFGs and two enhanced variants of DFGs (with additional arcs or rectangles) are evaluated. Our empirical study provides strong support for making an effort (to discover and) to explicitly visualize history dependencies. A ProM plug-in generating such explicit visualization is described in this paper.