Trace Ordering Strategy for Incremental Process Discovery
Student: Emanuel Domnitsch
Title: Trace Ordering Strategy for Incremental Process Discovery
Supervisor: Daniel Schuster, M.Sc.
Examiners: Prof. Dr. Wil van der Aalst, Prof. Dr. Stefan Decker
The execution of business processes generates large amounts of event data that is stored in information systems. The extraction and analysis of process behavior contained in event data can lead to valuable insights and uncover how processes are executed in the real world. Incremental process discovery algorithms aim to gradually construct process models from event data. For this purpose, the observed process behavior is incrementally incorpo-rated into a process model by repairing nonconforming parts. In this thesis, we therefore investigate how different orderings of incrementally added traces, i.e., observed process behavior, influence the quality, i.e., recall and precision, of process models. We propose a framework for constructing trace ordering strategies that can be used as extensions to an incremental process discovery algorithm. Furthermore, we present instantiations of the framework based on multiple introduced trace ordering methods. We apply different trace ordering strategies on real-life event data using an incremental process discovery algorithm and evaluate the constructed process models. The results of the evaluation show that the quality of the constructed process models can be significantly improved by using different trace ordering strategies based on the proposed trace ordering methods.