Hierarchical performance analysis for process mining

Process mining techniques use event data from operational and software processes to discover process models, to check the conformance of predefined process models, and to extend such models with information about bottlenecks, decisions, and resource usage. In recent years, the process mining field made huge advances in terms of scalability. In addition, recent work in process discovery supports advanced process model constructs such as subprocesses, recursive structures, cancellation, and various notions of concurrency. Hence, one has to realize that a simple, small, and flat model will not suffice anymore, especially when applied to analyzing software system processes. However, state of the art performance analysis is still typically performed either over the whole process model or at the level of individual activities. There is a lack of formal support for performance analysis on various submodel abstractions while taking into account the execution semantics. This paper presents 1) a framework for establishing precise relationships between events and submodels, taking into account execution semantics; and 2) a novel formalization of existing and novel performance metrics. Our approach enables advanced performance analysis at various submodel abstractions. An implementation is made available, and we demonstrate the advantages of our approach to various software system processes, showing the applicability and advantage with respect to existing techniques.