The PA group is conducting research, development, and experiments in the area of Process Analytics and Process Mining. If you have been matched to the PA group, you can join these endeavors by developing your own Process Analytics or Process Mining solution. We offer challenging assignments on industrially relevant topics, personal guidance developing your own cutting-edge solution in concept and in realization, and evaluation and experiments on real-life data.
The following assignments are a selection of possible topics and available at all times for students that have been matched to the PA group. If you have been matched and have a question about a particular topic, please contact the lecturer offering the topic. If you have ideas for your own topic, please contact Dr. Dirk Fahland.
Process Mining is increasingly shifting to graph-based representations of event data, enabling the application of Graph Neural Networks (GNNs) for various process mining tasks. Recent Master projects developed the first GNN-based process discovery algorithm which trains a GNN to translate an event log into a process model. The GNN achieves state-of-the-art performance in terms of Read More …
Industrial practice requires process mining techniques to not just produce models, dashboard, and data visualizations, but to help analysts get insightful answers to relevant questions. The emerging paradigm of object-centric process mining allows a new way to generate such insightful answers by embedding process mining results in the original domain data and context where the Read More …
The process mining field is exploring a new data model for event data called “object-centric event data” (OCED). In this data model, events are not partitioned under a unique case identifier, but each event is related to a number of data objects that can also be related to each other. The resulting data model essentially Read More …
The process mining field recently has adopted a graph-based approach for modeling and reasoning over event data using labeled property graphs. From a process mining perspective, events are related to the various objects and entities involved in a process, providing a more realistic description of actual process dynamics in relation to domain concepts. From knowledge Read More …
Highlights Background Business process management has evolved from manual workflow documentation to sophisticated data-driven systems that leverage machine learning for process optimization, anomaly detection, and predictive monitoring. Organizations across healthcare, finance, manufacturing, and logistics generate massive amounts of event logs that capture detailed execution traces of their operational processes. While deep learning approaches have shown remarkable success Read More …
Background Modern football analysis has undergone significant transformations with the advent of advanced data collection methods and artificial intelligence technologies. While traditional analysis relied solely on event data (passes, shots, tackles), the industry has shifted toward tracking data that captures player and ball positions throughout matches. Both high-quality tracking data and broadcast footage with sufficient Read More …
Have you ever analyzed some event data and wondered if the steps you take impact the results you get?Or whether the result you obtain match what you expected to find? These are common challenges faced by process mining analysts, who analyze large sets of event data to gain insights into how business processes are executed. Read More …
Process models are used to describe and reason about the execution of a process (e.g., package delivery) where a process instance (package), also called as a case, moves through the system. A case in a process is often subject to interaction with other cases and/or resources (e.g. deliverer), impacting the workflow. Event logs record which Read More …
Responsible Process Mining as a topic is introduced in [1]: “The prospect of data misuses negatively affecting our life has led to the concept of responsible data science. It advocates for responsibility to be built, by design, into data management, data analysis, and algorithmic decision-making techniques such that it is made difficult or even impossible Read More …
CPN Tools is a tool that is well-known in the Petri net community. CPN Tools provides a mature environment for constructing, simulating, and performing analysis of CPN (Coloured Petri Net) models. CPN Tools consists of an ML-based CPN simulator (the back-end), and a CPN editor (the front-end) that has been developed in the BETA programming Read More …
Smart cities approach does not only emphasize the implementations of new technologies in a city but also highlights the importance of using new technologies for enabling citizens’ engagement in urban planning processes. In that regard, ICTs play a vital role in (i) supporting citizens to report their complaints related to the public spaces (i.e. Read More …
StrProMCDD is a recently published work that detects concept drifts in event streams (see the figure below). StrProMCDD uses several model-based distance measures to detect these deviations using an adaptive window concept. In this assignment, we would like to compare the performance of this model-based approach with log-based stream clustering approaches that try to detect drifts in Read More …
A recent visual analytics technique called the “Performance Spectrum” https://github.com/processmining-in-logistics/psm allows us to gain more fine-grained insights into performance behavior and changes over time. A TU/e Master student showed that it is possible to mine synchronization of cases from the performance spectrum data showing that the behavior of a case depends on the mechanisms and Read More …
In many application domains, a process execution is captured using natural language. Think of medical records, customer complaints, legal records… The same holds for process models: they can be captured as text for medical guidelines, user manuals, legal regulations are typical examples of such cases. Such data forms a new challenge for the process mining Read More …
Available process discovery have been tested in the customer journey context under offline settings. Recent online process discovery approaches like: https://ieeexplore.ieee.org/document/7376771 bring however a lot of added value for a real-time customer journey optimization. The objective of this assignment is to use two different customer journey datasets to test the effectiveness of such approaches for Read More …
The analysis of the temporal evolution of dynamic graphs like social networks is a key challenge for understanding complex processes hidden in graph structured data. Graph evolution rules capture such processes on the level of small subgraphs by describing frequently occurring structural changes within a network. Existing rule discovery methods make restrictive assumptions on the Read More …
BFSPMiner is an effective and efficient batch-free algorithm for mining sequential patterns over data streams was published very recently https://link.springer.com/article/10.1007/s41060-017-0084-8. An implementation of the algorithm is available here: https://github.com/Xsea/BFSPMiner. As BFSPMiner has proven to be effective (see Figures 10-14 of the paper) in different domains (see Table 1 in the paper), we would like to Read More …
In the context of process mining, we are often confronted with companies willing to share their data if we can sufficiently anonymize this. However, to date, there are no well-defined plugins to do such anonymizations. Therefore, we are looking for a Master student that is willing to help us with this. Part of the project Read More …
Aligning structured process models to event logs is a far from trivial task. In complex modelling languages, inclusive OR-split/join patterns play an important role and they are known to be notoriously difficult to align to event logs due to their large state-spaces. The known Petri net translations of OR-joins rely either on token coloring or Read More …
For experimenting with process discovery and Petri nets, scientists often rely on experiments with artificial models and logs. More often than not, these models are block structured as it is easy to generate such models by simply building a random process tree and translating that into a Petri net. However, Petri nets allow for more Read More …
Replaying event logs on Petri nets, either through token-replay or using alignments, is a complex task. Especially when models become larger and have more labels, the size of the models becomes a problem. In Petri net theory, many reduction rules exist for reducing Petri nets while retaining, for example, soundness of the model. Can we Read More …
