Augmented Change Point Detection in Process Mining by Incorporating Time, Data, and Resource Information
MSc Thesis Project
Title: Augmented Change Point Detection in Process Mining by Incorporating Time, Data, and Resource Information
Author: Jan Niklas Adams
1st examiner: Prof. Dr. Thomas Rose
2nd examiner: Prof.Dr.ir Wil M.P. van der Aalst
The analysis of business processes can yield huge benefits to corporations, potentially saving high amounts of money and increasing the customer satisfaction. Since significant changes in a business process often have an impact on either the cost or the customer satisfaction, it is beneficial to detect the changes and uncover the underlying cause-effect relationships. This knowledge can help to improve the process and predict the future behaviour of the process. In this thesis, we therefore propose a framework, that searches for concept drifts in different perspectives of a business process and tests, whether a cause effect relationship between these drifts is present. We define methods for transforming an event log into a time series representation of different process perspectives, detecting change points in this time series and testing for cause-effect relationships. We evaluate this framework using a synthetic event log and a real-life event log, that both contain cause-effect relationships between different perspectives. We can demonstrate, that our framework can indeed uncover cause-effect relationships and provide recommendations for the user of this framework.