[1/5] How do we make object-centric data ingestion flexible, scalable, and maintainable? Our latest collaboration (with Lien Bosmans, Alexandre Goossens, @Gio_Lugaresi, @Johannnesss, and @jochendw ) discusses this from a data-engineering perspective.

The paper is already online: (5/5).

Additionally, we explore the use of adjusted and customizable loss functions that allow end-users to find the optimal independence-predictive performance trade-off for their specific problem. (4/5).

More specifically we explored using metrics for demographic parity, such as ΔDP , as well as recently introduced, threshold-independent distribution-based alternatives. (3/5).

In this work we addressed group fairness in predictive process monitoring by investigating independence, aligning predictive models with legal and ethical standards. (2/5).

Happy to be in Vienna this week for the CAiSE Conference! . 📣 Catch our talk on Wednesday at 16:00 titled "Achieving Group Fairness through Independence in Predictive Process Monitoring", work in collaboration with Simon De Vos. (1/5).

RT @paoloceravolo: Process Model Forecasting: Univariate vs Multivariate Approaches. Yongbo Yu, Jari Peeperkorn, Johannes De Smedt and Joch….

[5/5] And which also contains other functionalities such as automated data quality tests , on-demand data extraction to various event log formats, and visualisation for interactive data exploration. The repository can be found here:.

[4/5] We also introduce the Stack't tool, an end-to-end solution that implements the relational schema using a lightweight, open-source data stack, which is able to efficiently ingest process data from different sources.

[3/5] We then introduce a hub-and-spoke architecture to decouple data sources from process mining applications, and an accompanying relational schema, robust against changes in the process data archtecture, and compatible with existing event log formats (such as OCEL 2.0).

[2/5] In this work we propose a set of requirements that robust object-centric event log storage, ingestion, and preparation should ideally adhere to. With a focus on among others dynamic environments, functionality, and robustness.

RT @Eurosport_NL: 🔥🇫🇷 | Worthy de Jong maakt de tweepunter. De vlaggendrager ziet zijn olympische droom uitkomen. Nederland heeft GOUD! 🤩🥇….

RT @jochendw: @EstefaniaSerral @Macuyiko And finally a non-LIRIS @FEBkuleuven contribution to close this session, by DTU's @lopez_handres t….

RT @jochendw: @EstefaniaSerral Third up is Brecht Wuyts explaining his newest transformer model for suffix prediction in PPM. Research in c….

RT @jochendw: Second presentation by Yannis Bertrand on trace clustering and explainability. Work together with @EstefaniaSerral https://t.….

RT @jochendw: Tuesday at the @k_conference and another session on the topic of process mining. First up, Jakob De Moor presenting his work….

RT @jochendw: And also @Jari_i presented research led by @APMJStevens on counterfactuals for OOPPM based on manifold learning. Great sessio….

RT @jochendw: Great to also have @tthildebrandt presenting his work on DCR graphs for XAI during our session @EUROonline_News conference. @….

RT @jochendw: Andrei Buliga explaining his research on XAI in predictive process monitoring at the @EUROonline_News conference, within the….

RT @jochendw: To all, especially early stage, PhD researchers in process mining, join @hanvanderaa and myself in Kopenhagen for a day full….