Causal Process Mining: Concepts and Tool

WU

Process mining has struggled with so called Spaghetti models ever since its invention. So far, most approaches try to address this problem by abstraction and filtering. In this project, we argue that the problem behind Spaghetti models is fundamentally rooted in the semantic relationships that most process mining algorithms use. Instead of the causality between events, algorithms works with directly-follows relations which leads to spurious relations. Based on this observation, the goal of this project is to define a novel mining approach that considers casual relation between events. Its key idea is to first semantically enrich observed execution sequences before aggregating them. Our proposed foundational research project includes the development of novel concepts and algorithms, their validation, and the development of a tool for practical use and experimental validation.

The research described by this proposal will follow the design science methodology ~\cite{peffers2007design}. We organized the envisioned contribution in three work packages (WPs). WP1 is concerned with the design of the causal process mining algorithms. WP2 is concerned with the development of a prototype that implements the concepts and algorithms for causal process mining. WP3 focuses on the evaluation of the causal process mining algorithms.

With a strong background in both BPM and Data mining, the applicant of this project, Dr. Kate Revoredo, is ideally qualified for leading this research. She is already collaborating with researcher Philipp Waibel on the topic of the project which will allow for a quick acceleration of the outlined project.