Business Process Mining in der Schweizerischen Nationalbank: Erarbeitung eines theoretisch hergeleiteten und empirisch validierten Bezugsrahmens zur Daten- und Ereignisprotokollqualität

Business Process Mining (BPM) allows reconstructing and analyzing process models based on log data and may become important for the Swiss National Bank (SNB) when implementing future strategic and operational initiatives.
The author reviews the relevant literature which shows that BPM has been widely discussed since the start of the century. A key factor for the successful mining of reliable process models is high data quality. The main objective of this thesis is to use expert interviews to develop practical solutions to data-specific problems shown in an SNB case study analysis with real payment transaction data that has not yet been sufficiently covered in the academic literature. Preventive measures to improve the quality of data and event logs are also discussed.
The results of the analysis show that the quality of the available data is appropriate for applying BPM to operational processes in the core banking platform "Avaloq Banking System". A number of data-specific issues, however, become apparent: (1) The timestamp shown in the event log does not reflect the actual time of the activity; (2) the characteristics of individual attributes are partly only available in free text format; (3) individual workflow actions have consistent main designations but different modifiers and (4) certain events occur in reality but are not recorded in the event log. From a practical point of view, there are various solutions to the issues (1) - (4): Solutions include a change of the system logging logic (issue 1); the use of robotics in combination with artificial intelligence to create attribute names and define categories (issue 2); applying ontology to create a relationship between workflow actions (issue 3) and the use of workflow engines (issue 4). One final result is that data quality for the mining application can also be approached preventively, e.g. by performing past based error analyses at system field level.
This dissertation provides both theoretical and practical insights and includes a solid understanding of the analyzed business process and a comprehensive overview of the literature regarding data quality problems and solutions.