A Comparative User Study on Control-Flow Visualizations of Event Log Data

2026-06-30

Sanne van der Linden *

Martijn C. Willemsen *

M. D. Brunings *

Anna Vilanova *

Stef van den Elzen *

DOI：10.1111/cgf.70510

Original

Save

View PDF

Abstract

En 中文

Event sequence data is collected abundantly, especially in the Process Mining (PM) and Visual Analytics (VA) fields. Despite these two fields occasionally working on similar challenges, there is little cross-fertilization. While these two fields can complement each other based on their strengths, both fields have independently developed abstractions and visualizations to understand and analyze the control-flow of event sequences. However, there is no study that evaluates the strengths and weaknesses of these abstractions and visualizations for different common tasks. This paper compares seven commonly used control-flow visualizations using a formal user study with 48 participants. Directly-Follows Graphs (DFGs), Petri nets, Business Process Model and Notations, Process trees, variant timelines, Icicle plots, and Sankey diagrams are evaluated on their performance on different tasks, understandability, and user preferences. Our results show that the DFG, variant timeline, and Sankey diagram are the only visualizations that significantly outperform others for the selected tasks. Furthermore, DFG significantly outperforms most others in identifying the most frequent activity. Surprisingly, DFG also excels in identifying loops of three consecutive events despite being hard to answer correctly. Despite its high performance, we find that the DFG is not preferred by participants and even more so when participants do not understand it. Moreover, we find that understandability influences user preferences and correctness.

Keywords:

process mining

user study

visual analytics

A Comparative User Study on Control-Flow Visualizations of Event Log Data

Abstract

AI Summary

Journal

Organization

Cited Papers

Citing Papers