Foundations of Streaming Process Mining - From Representation to Evaluation Discussions at SMA4PM 2025

Zenodo (CERN European Organization for Nuclear Research)

Abstract

Streaming process mining is an emerging research area at the intersection of process mining and data stream management. Unlike traditional process mining, which assumes access to complete event logs, streaming scenarios require methods and representations that can cope with continuous, potentially infinite flows of events, where cases may remain incomplete and the system may evolve over time. This poses new conceptual, technical, and evaluation challenges, ranging from the minimal representation of processes under uncertainty, to algorithmic paradigms that work under memory and latency constraints, to suitable benchmarks for assessing quality and reproducibility. This document summarizes the discussions and open challenges across two key themes: conceptual models and challenges for streaming process mining, and techniques and benchmarks for streaming process mining.

Authors 9

1. Andrea Burattin

   Technical University of Denmark

   Affiliation as printed

   Technical University of Denmark

2. Gabriel Marques Tavares

   Ludwig-Maximilians-Universität München

   Affiliation as printed

   Ludwig-Maximilians-Universität München

3. Jeppe Mikkelsen

   Technical University of Denmark

   Affiliation as printed

   Technical University of Denmark

4. István Koren

   Eötvös Loránd University

   Affiliation as printed

   Eötvös Loránd University

5. Andrea Maldonado

   Technical University of Munich

   Affiliation as printed

   Technical University of Munich

6. Christian Imenkamp

   University of Bayreuth

   Affiliation as printed

   University of Bayreuth

7. Lukas Liß Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University

8. Miriam Wagner Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University

9. Bhavin Kamani
   Affiliation as printed

   Gembatech