Beyond Log and Model Moves in Conformance Checking: Discovering Process-Level Deviation Patterns

Process managers apply conformance checking techniques to identify deviations between the desired and the actual execution of a process. From a process-level perspective, these deviations often involve multiple interrelated events, for example if activities are executed in the wrong order or are unnecessarily repeated. However, state-of-the-art conformance checking techniques do not reveal these process-level deviations, instead identifying only event-level deviations in the form of inserted or skipped events. To address this shortcoming, this paper presents an approach that discovers process-level deviations from event-level insights provided by alignment-based conformance checking techniques. These deviations are discovered as instantiations of five commonly used patterns of non-conformance: inserted, skipped, repeated, replaced, and swapped. The approach is designed to choose patterns according to a user’s preferences and contextualize them within parallelism and choices in the process model. Our evaluation shows that it reliably detects process-level deviations, thus providing process managers with more comprehensive information on process conformance.