back to list

Project: Interactive Process Discovery

Description

Process discovery is often done automatically using process discovery algorithms, such as the Probabilistic Inductive Miner [2]. However, the quality of the discovered models depends heavily on the quality of the data and how the data is prepared and selected. This is hard to do automatically. On the other hand, the user often has a lot of domain knowledge on the process that cannot be found in the data. We want to investigate how we can, using visual analytics and the combination of data and user’s domain knowledge, enable the user to investigate the data set (event log) characteristics to build better models using a combination of visualization, interaction, and automated methods.

The project is twofold, first we want to investigate how the process discovery algorithm can be made interactive such that users can use their domain knowledge to make potentially better decisions for the resulting model, and second, we want to improve the resulting visualization of the model (e.g., to prevent ambiguations).

Requirements

•        Good programming skills
•         Understanding of process mining models
•        Visualization background

References

[2] D. Brons, R. Scheepens and D. Fahland, "Striking a new Balance in Accuracy and Simplicity with the Probabilistic Inductive Miner," 2021 3rd International Conference on Process Mining (ICPM), Eindhoven, Netherlands, 2021, pp. 32-39, doi: 10.1109/ICPM53251.2021.9576864.

Details

Student

WH

Wouter van der Heijden

Supervisor

Stef van den Elzen