Event details from: Business Analytics Seminar: Matthijs van Leeuwen (Leiden University)

Date
16 March 2026
Time
12:00 – 13:00

Abstract:

Interpretable machine learning is becoming more and more important, as many data scientists and domain experts have started to realise that the use of large 'black-box' models is not always desirable or even possible. For example, in health care there is often a need for data-driven decision support, but this must be fully transparent in order to be trustworthy to medical doctors. Rule set learning might be able to fill this gap, and has been frequently revisited because of its inherent interpretability.

Existing rule learning methods have several shortcomings though. First, most existing methods impose orders among rules, either explicitly or implicitly, which makes the models less comprehensible. Second, due to the difficulty of handling conflicts caused by overlaps (i.e., instances covered by multiple rules), existing methods often do not consider probabilistic rules. Third, learning classification rules for multi-class target is understudied, as most existing methods focus on binary classification or multi-class classification via the ``one-versus-rest" approach.

To address these shortcomings, we proposed TURS, for Truly Unordered Rule Sets. To resolve conflicts caused by overlapping rules, we propose a model that exploits the probabilistic properties of our rule sets, with the intuition of only allowing rules to overlap if they have similar probabilistic outputs. We next formalise the problem of learning a TURS model based on the minimum description length principle and carefully design a heuristic algorithm. We benchmark against a wide range of rule-based methods and demonstrate that our method learns rule sets that have lower model complexity and highly competitive predictive performance. Finally, we apply TURS to patient data collected at an intensive care unit (ICU), and identify patient subgroups with deviating readmission risks and show that these subgroups are characterised by heterogeneous feature distributions.

Bio:

Matthijs is associate professor, group leader of the [Explanatory Data Analysis](https://eda.liacs.nl/) group, and director of education at the [Leiden Institute of Advanced Computer Science](https://liacs.leidenuniv.nl/) (LIACS), the computer science and AI institute of Leiden University. His primary research interest is exploratory data mining: how can we enable domain experts to explore and analyse their data, to discover structure and---ultimately---novel knowledge? He was previously a tenure track assistant professor (2017-2020) and senior researcher (2015-2017) at Leiden University, and a postdoctoral researcher at KU Leuven (2011-2015) and Universiteit Utrecht (2009-2011). He defended his Ph.D. thesis, titled 'Patterns that Matter', at Universiteit Utrecht in February 2010. He won several best paper awards at international conferences and was awarded NWO Rubicon, FWO Postdoc, and NWO TOP2 grants. He is General Chair of the IDA Council and Action Editor of the Data Mining and Knowledge Discovery journal. Further, he co-organised a number of international conferences and workshops, and he was guest editor of a TKDD special issue on ‘Interactive Data Exploration and Analytics’.

General information:

This seminar will be organised in a hybrid setup. If you are interested in joining this seminar, please send an email to the secretariat of Amsterdam Business School at secbs-abs@uva.nl.