A Generic Approach Towards Feature Extraction from Object-Centric Event Logs

 

Student: Johannes Herforth (MSc)

Title: A Generic Approach Towards Feature Extraction from Object-Centric Event Logs

Supervisor: Mahnaz Qafari, Alessandro Berti

1st Examiner: Prof. Wil M.P. van der Aalst

2nd Examiner: Prof. Ulrik Schroeder

Summary

The object-centric event log standard is a novel data type which allows for multiple case notations to be represented in the same document. Therefore, by applying feature extraction, it is possible to gain knowledge of how the case notations interact with each other. Though, extrapolating how and why the different case-notations interact and representing it in terms of features is not a trivial task. The goal is to create methods to extract generic features for use in data analysis or machine learning problems. To be useful, the features must be easy to understand, scalable and perform well in data analytical tasks. This paper proposes a diverse set of methods to represent object-centric event logs and extract generic features based on point, local and global views of objects and events. The evaluation is based on solving data analytical tasks on top of a simulated object-centric log using the proposed methods. While the methods do show positive results in the evaluation, many are also highly dependent on the choices done during the OCEL extraction.