Human decision point mining in semi-automated workflows

Abstract

Within the domain of process mining, automated business rule discovery can greatly advance the discovery and conformance checking of business rules. Nevertheless, the human applied business rules and decision points in a business process remain subjective and can differ greatly between employees, between departments, and between automated rules. This can leave a gap in the overall quality of business rule understanding and governance. Therefore, we present a mining approach to reveal the human decision points within a set of mined decision points. This approach could support process owners and management in revealing the hard-to-find business logic from human actors. Our research is based on a behavior detection technique from online gaming (Choi, 2016), and in 5 consecutive process mining steps, we aim to provide a multi-perspective model that high lights the human decision points. We applied our proposition onto an event log from an unstructured semi-automated workflow process in the process mining tool ProM (ProM, 2019). The results show progress in the individual five mining steps but errors in intermediate model compatibility and comparison. Although individual working ProM implementations deliver sound intermediate models, we did not succeed to create a stable mining experiment to demonstrate our proposal. Further research is needed to define better compatible models and explore other supporting tooling, more tailored to data mining, to create a better behavior model. A richer data set will increase the chance of having more human behavior distinguishing attributes.

Date of Award	21 Jun 2020
Original language	English
Supervisor	Lloyd Rutledge (Examiner) & Stef Joosten (Co-assessor)