On the Use of Steady-State Detection for Process Mining: Achieving More Accurate Insights

On the Use of Steady-State Detection for Process Mining: Achieving More Accurate Insights

Abstract

Steady-state detection (SSD) is a critical task in the analysis of dynamic systems, as it enables the reliable evaluation of system behavior by differentiating between stable and unstable states. While SSD techniques have been developed and tested in domains such as signal processing and industrial systems, their application in the information systems domain, particularly in process mining, has been largely overlooked. Specifically, event logs that record the executed behavior of a business process often contain data from both steady and non-steady states, which can distort process mining results, such as performance analysis and remaining time prediction. This paper highlights the importance of SSD in the process mining domain and investigates the applicability of existing SSD solutions. To operationalize this, we propose a two-step framework for detecting steady states in business processes. The framework extracts relevant process characteristics from an event log and applies established SSD techniques to identify periods during which a business process operated in a steady state. We evaluate the framework through experiments that assess its accuracy within a controlled environment using simulated event logs and that demonstrate the benefits of SSD for a downstream process mining task: remaining time prediction. The findings emphasize the potential of SSD for obtaining more accurate process mining insights.

Grafik Top
Authors
  • Kraus, Alexander
  • Elyasi, Keyvan Amiri
  • van der Aa, Han
Grafik Top
Shortfacts
Category
Paper in Conference Proceedings or in Workshop Proceedings (Paper)
Event Title
Advanced Information Systems Engineering
Divisions
Workflow Systems and Technology
Subjects
Informatik Allgemeines
Event Location
Vienna, Austria
Event Type
Conference
Event Dates
June 16-20, 2025
Publisher
Springer Nature Switzerland
Page Range
pp. 204-220
Date
2025
Export
Grafik Top