This thesis introduces action-oriented process mining, a novel approach that closes the critical gap between process diagnosis and improvement implementation that has long challenged organizations. Unlike traditional process mining that stops at diagnostics, this research establishes the first comprehensive framework that transforms datadriven insights into concrete actions through three integrated components: object-centric problem monitoring, pattern-based action generation, and data-driven impact evaluation. Implemented as the open-source ProAct web application, this framework demonstrates both theoretical significance and practical applicability, setting a new direction for the BPM field.
Eforts to improve business processes have evolved from methodologies like business process
reengineering, which aimed for radical changes, to approaches like process redesign and lean management
addressing incremental improvements [
Process mining emerged as a response to these limitations, ofering a data-driven approach that
analyzes real operational data extracted from information systems [
The framework for action-oriented process mining [
CEUR
ceur-ws.org
Process-Aware Information Systems
Event data
Process monitoring
Problem instances
Action engine
Action instances
Impact analysis
Impacts change/update define problems/actions process managers adaptations, resource reallocation, business rule modifications, or implementing alerts [ 13, 14]. After implementation, the impact analysis component evaluates the efects of these actions by analyzing new event data, measuring the structural, operational, and performance impacts [15]. This analysis provides critical feedback to process managers, enabling data-driven decisions for continuous improvement and completing the process improvement cycle.
Building upon the action-oriented process mining framework described in the background section, this thesis addresses four fundamental research challenges: • RQ1: How can process monitoring techniques accurately identify operational problems in complex, real-life business processes? The first component of the framework, i.e., process monitoring, faces significant challenges when applied to real-world scenarios. Existing approaches often assume simplistic, single-case processes that fail to capture the dynamic and object-centric nature of real-life business processes where multiple interconnected objects flow through processes simultaneously. • RQ2: How can an action engine systematically analyze temporal patterns of operational problems to generate contextually relevant actions? Once problem instances are identified by the monitoring component, the action engine must translate these insights into concrete improvements. This question explores how to automate the generation of appropriate actions by analyzing temporal patterns of operational problems rather than treating them as isolated incidents, addressing the second key component of the framework. • RQ3: How can comprehensive impact analysis techniques be developed to evaluate the structural, operational, and performance efects of actions implemented for process improvement? This question investigates how to systematically analyze the impacts of process updates, i.e., actions, on diferent organizational aspects, including the structural elements of a process, the operational state, and performance indicators. • RQ4: How can a robust simulation approach be developed that accurately emulates the complexities of real-world processes to efectively evaluate action-oriented process mining techniques? To validate the framework before operational deployment, we need simulation environments that reliably reproduce real-world complexities. This question examines how to create such environments by incorporating authentic system data and configurations.
Note that RQ1-3 address the development of eficient techniques for implementing the three main components of the action-oriented process mining framework introduced in the background section, whereas RQ4 concerns providing an efective testbed for developing and evaluating these techniques.
The thesis makes the following key contributions, each directly addressing the research questions and advancing the components of the action-oriented process mining framework:
Beyond traditional, single-case process mining, this contribution presents a comprehensive taxonomy of object-centric operational problems that captures the complexity of processes involving multiple interconnected objects. The taxonomy classifies problems into compliance-oriented and performanceoriented categories, providing a foundation for more accurate problem detection. Object-centric problem graphs with formally defined semantics are introduced to visually represent these problems, enabling process managers to define complex operational issues [ 16, 17]. The developed monitoring engine analyzes object-centric event logs against these problem definitions, surpassing traditional single-case monitoring methods by providing a more realistic view of operational issues in real-life object-centric processes.
Moving from insights to actionable improvements, a pattern-based approach is developed that systematically transforms problem instances into concrete interventions [18]. A comprehensive taxonomy of actions for process improvement provides a structured way to address diferent types of operational problems with clear semantics. The action graphs visually represent problem-action relationships, helping process managers understand recommended interventions. By analyzing temporal dependencies among problem instances, the engine produces conflict-free action plans that resolve identified issues while respecting execution constraints, providing clear implementation guidelines for process managers.
Completing the improvement cycle, a comprehensive approach [19] using Digital Twin Interface Models (DT-IM) [20] is presented. This method identifies changes caused by action instances and performs three distinct analyses: 1) structural impact analysis to identify afected activities and business functions, 2) operational impact analysis to determine afected process instances and business objects, and 3) performance impact analysis to measure changes in key performance indicators [21]. This data-driven approach provides valuable feedback to process managers, enabling informed decisions about future process improvements and completing the continuous improvement cycle outlined in the framework.
Bridging theory and practice, a novel simulation approach is developed that directly incorporates data and parameters from real-life information systems [22]. Rather than focusing solely on process flows, the simulation framework integrates the complexity of supporting information systems, including data lfows, business rules, and system configurations. By using actual system data, the approach efectively mirrors real-world complexities, creating a reliable testbed for validating all three components of the action-oriented process mining framework before operational deployment.
The contributions above have been implemented as an open-source web application, ProAct1, demonstrating both theoretical soundness and practical applicability. 1The user manual is available at https://proact.readthedocs.io/en/latest/. ProAct’s core analysis functionalities are implemented in python library OCPA [23].
This thesis represents the first comprehensive academic treatment of Action-Oriented Process Mining, introducing a general framework and presenting a suite of techniques that facilitate its implementation. These techniques address the critical challenges across the entire process improvement lifecycle from problem identification through action generation to impact evaluation.
Key innovations include: (1) object-centric process monitoring that accurately identifies operational problems in complex business processes, overcoming the limitations of traditional approaches; (2) a pattern-based action engine that systematically transforms insights into actionable improvements by analyzing temporal patterns and generating conflict-free action plans; (3) a data-driven impact analysis framework using Digital Twin Interface Models to provide detailed evaluation of implementation efects; and (4) a robust simulation approach creating realistic testing environments by incorporating real system data.
These contributions address the fundamental gap between process diagnosis and improvement implementation that has long challenged organizations. By providing a complete methodology that connects monitoring, action planning, and impact assessment, this thesis establishes a foundation for more efective and systematic process improvement. The open-source ProAct application implements these contributions, making them accessible to practitioners and researchers.
This research opens a new and important direction within the BPM field . Future work could focus on four key directions: (1) enhancing object-centric problem graphs to incorporate multiple perspectives; (2) integrating the action engine with leading workflow automation platforms; (3) conducting comprehensive case studies in diverse organizational settings; and (4) strategically connecting the framework with established process mining solutions to create a seamless ecosystem. These advancements will help transition this research from a theoretical framework to a practical solution capable of addressing real-world process improvement challenges across diverse organizational contexts.
The author acknowledge the use of Grammarly and Claude for grammar checking and language editing to improve the clarity of this manuscript. [9] G. Park, J. Benzin, W. M. P. van der Aalst, Detecting context-aware deviations in process executions, in: C. D. Ciccio, R. M. Dijkman, A. del-Río-Ortega, S. Rinderle-Ma (Eds.), BPM Forum, volume 458 of LNBIP, Springer, 2022, pp. 190–206. [10] G. Park, J. N. Adams, W. M. P. van der Aalst, Conformance checking and performance analysis using object-centric directly-follows graphs, in: A. Marrella, M. Resinas, M. Jans, M. Rosemann (Eds.), BPM Forum, volume 526 of LNBIP, Springer, 2024, pp. 179–196. [11] J. N. Adams, E. Hastrup-Kiil, G. Park, W. M. P. van der Aalst, Super variants, in: A. Marrella,
M. Resinas, M. Jans, M. Rosemann (Eds.), BPM, volume 14940 of LNCS, Springer, 2024, pp. 111–128. [12] G. Park, W. M. P. van der Aalst, Action-oriented process mining: bridging the gap between insights and actions, Progress in Artificial Intelligence (2022). [13] M. Reichert, B. Weber, Enabling Flexibility in Process-Aware Information Systems - Challenges,
Methods, Technologies, Springer, 2012. [14] G. Park, M. Song, Optimizing resource allocation based on predictive process monitoring, IEEE
Access 11 (2023) 38309–38323. [15] M. Parhizkar, M. Comuzzi, Impact analysis of ERP post-implementation modifications: Design, tool support and evaluation, Comput. Ind. 84 (2017) 25–38. [16] G. Park, W. M. van der Aalst, Operational process monitoring: An object-centric approach,
Computers in Industry 164 (2025) 104170. [17] G. Park, W. M. P. van der Aalst, Monitoring constraints in business processes using object-centric constraint graphs, in: M. Montali, A. Senderovich, M. Weidlich (Eds.), ICPM Workshops, volume 468 of LNBIP, Springer, 2022, pp. 479–492. [18] G. Park, D. Schuster, W. M. P. van der Aalst, Pattern-based action engine: Generating process management actions using temporal patterns of process-centric problems, Comput. Ind. 153 (2023) 104020. [19] G. Park, M. Comuzzi, W. M. P. van der Aalst, Analyzing process-aware information system updates using digital twins of organizations, in: R. S. S. Guizzardi, J. Ralyté, X. Franch (Eds.), RCIS, volume 446 of LNBIP, Springer, 2022, pp. 159–176. [20] G. Park, W. M. P. van der Aalst, Realizing A digital twin of an organization using action-oriented process mining, in: C. D. Ciccio, C. D. Francescomarino, P. Sofer (Eds.), ICPM, IEEE, 2021, pp. 104–111. [21] G. Park, J. N. Adams, W. M. P. van der Aalst, Opera: Object-centric performance analysis, in: J. Ralyté, S. Chakravarthy, M. K. Mohania, M. A. Jeusfeld, K. Karlapalem (Eds.), ER, volume 13607 of LNCS, Springer, 2022, pp. 281–292. [22] G. Park, W. M. P. van der Aalst, Towards reliable business process simulation: A framework to integrate ERP systems, in: A. Augusto, A. Gill, S. Nurcan, I. Reinhartz-Berger, R. Schmidt, J. Zdravkovic (Eds.), BPMDS, volume 421 of LNBIP, Springer, 2021, pp. 112–127. [23] J. N. Adams, G. Park, W. M. P. van der Aalst, ocpa: A python library for object-centric process analysis, Softw. Impacts 14 (2022) 100438.