The Research Group Business Informatics, or onderzoeksgroep Beleidsinformatica in Dutch, is part of the department Quantitative Methods of the Faculty of Business Economics at Hasselt University. Our group conducts research to support businesses and organisations in decision-making by modelling data, information and knowledge. The main focus points are process analytics in healthcare, behavioural analytics, digital audit analytics and explainable AI.
With so many different IT systems capturing and registering data, very rich information about the actual execution of business processes is available. Unfortunately, data at such a fine-grained level brings many challenges to both business users and analysts. On the one hand, there may be a mismatch between the level at which data is registered and the mental model about a process from the business user. On the other hand, analysts find it complex to gather concise insights from the data also due to the large quantity that is available for analysis. A good understanding of both the process and registered data is therefore essential for analysts. Of course, this process notion does not have to be present. A server hosting IT applications can also register event data on a fine granularity.
The first issue, the fine-grained registration of data, can be tackled with event log abstraction: taking this low-level data and increasing the granularity level. Several abstraction techniques exist in the field already, yet they sometimes rely on domain knowledge.
A fully automated approach that produces high-quality results is still a topic heavily under investigation. One such approach is transforming the event log by imputing event abstraction patterns, which can be automatically learned from the data.
After obtaining this better understanding of the data by increasing the abstraction level, my goal is to aid analysts in executing a causal analysis. A key differentiation here is performing a causal analysis instead of a correlation analysis, which is most often presented as a causal analysis. We, therefore, also take a deeper dive into the underlying philosophies about the definitions of causal relations.
As such, my Ph.D. research can be summarised as enabling a fast, convenient, and accurate diagnostic analysis in a data-driven fashion for low-level event data.
I am doing my PhD in the Research Group Business Informatics under the supervision of prof. dr. Benoît Depaire and prof. dr. Niels Martin.
In this more and more digital world, data becomes more and more fine-grained. My solution to trying to tackle this, and thus making automatically learned process models via process discovery more comprehensible, is unsupervised event log abstraction. More specifically, I have worked around using event abstraction patterns to replace low-level events with high-level activities without domain expertise. My research thus far has shown that these techniques do not accomplish this task without sacrificing the representativeness of the underlying low-level data: fitness and precision generally take heavy hits.
Current and future work includes expanding the already performed experiment to acquire more generalisable results. Also, a collaboration with the University of Padua (IT) is being set up to see which solution we can propose to the research community.
Determining the root causes of, for example, negative case outcomes, is vital for businesses to improve. However, effective causal analysis is much more meaningful than a correlation analysis. Unfortunately, most recent works sell a correlation analysis as if it is representing causal relations. To that end, we take a deep dive into the underlying philosophies regarding causality theory. What defines a causal relation and can we reach a consensus? Business users want to be sure that, what is pointed at as the root cause of an issue, is also the root cause. Our first step in that sense was proposing a new algorithm, AITIA-PM, that takes a probabilistic approach in identifying root causes while controlling for confounding factors.
Current and future work includes providing additional support for the PTL language in defining the causal hypotheses and supporting continuous variables.
The table underneath shows an overview of my publications thus far. When available, a button is included to download the author's version of the paper if desired. Additional details can be found on my researcher profile on the UHasselt DocumentServer. You can also visit my ORCID ID and Google Scholar profile.
Reference | Category | Manuscript |
Nápoles, G., Van Houdt, G., Laghmouch, M., Goossens, W., Moesen, Q., & Depaire, B. (2020). Fuzzy cognitive maps: A business intelligence discussion. InIntelligent Decision Technologies 2019(pp. 89-98). Springer, Singapore. | ||
Van Houdt, G., Mosquera, C. & Nápoles, G. (2020). A review on the long short-term memory model. Artif Intell Rev 53, 5929–5955 (2020). | A1 | |
Martin, N., Van Houdt, G., Janssenswillen, G. (2020). Towards more structured data quality assessment in the process mining field: the DaQAPO package. In: Book of abstract for European R Users Meeting 2020. | C2 | |
Van Houdt, G. (2020). Mining Behavioural Patterns from Event Data to Enable Context-Aware Root Cause Analysis. In: Di Ciccio, Claudio; Depaire, Benoît; De Weerdt, Jochen; Di Francescomarino, Chiara; Munoz-Gama, Jorge (Ed.). In: Proceedings of the ICPM Doctoral Consortium and Tool Demonstration Track 2020 co-located with the 2nd International Conference on Process Mining (ICPM 2020), p. 9 -10. | C1 | |
Van Houdt, G., Depaire, B., & Martin, N. (2020). Unsupervised Event Abstraction in a Process Mining Context: A Benchmark Study. In: ICPM Workshops (pp. 82-93). | C1 | |
Janssenswillen, G., Mannhardt, F., Creemers, M., Depaire, B., Jans, M., Jooken, L., Martin, N., Van Houdt, G. (2020). Extensions to the bupaR Ecosystem: An Overview. In: Proceedings of the ICPM Doctoral Consortium and Tool Demonstration Track 2020 co-located with the 2nd International Conference on Process Mining (ICPM 2020), p. 43-46. | C1 | |
Van Houdt, G., Depaire, B., Martin, N. (2022). Root Cause Analysis in Process Mining with Probabilistic Temporal Logic. In: Munoz-Gama, J., Lu, X. (eds) Process Mining Workshops. ICPM 2021. Lecture Notes in Business Information Processing, vol 433. Springer, Cham. | C1 | |
Martin, N., Van Houdt, G., & Janssenswillen, G. (2022). DaQAPO: Supporting flexible and fine-grained event log quality assessment. Expert Systems with Applications, 191, 116274. | A1 | |
Van Houdt, G., Martin, N., Depaire, B. (2023). AITIA-PM: Discovering the True Causes of Events in a Process Mining Context. (In Press) | A1 |
I have presented at the following conferences:
The 2nd International Conference on Process Mining (ICPM 2020):
www.icpmconference.org/2020 | www.edba.science
The 3rd International Conference on Process Mining (ICPM 2021):
www.icpmconference.org/2021 | www.edba.science
I was a member of the following Program Committees:
The 3rd International Conference on Process Mining (ICPM 2021):
The 4th International Conference on Process Mining (ICPM 2022):
Workshop on Event Data and Behavioral Analytics | Demo Track
The 5th International Conference on Process Mining (ICPM 2023):
Workshop on Event Data and Behavioral Analytics | Demo Track
Logo and design by Felita.
