Process Comparison Using Petri Net Decomposition

Business processes drive the value creation at companies requiring them to constantly monitor and improve the former. The field of Process Comparison (PC) offers promising approaches to gain insight into differences between variants of a process that one can leverage to improve the latter. For example, one might consider the same process at different points in time or at different sites. Recent PC methods consider event logs containing data on real-life process executions the single source of truth. However, there often exist additional specifications that can be represented as Petri nets. In this paper, we propose an approach that leverages a given Petri net to compare two event logs in a hierarchical manner. To this end, we decompose the provided net into subprocesses and extract data on their executions from the event logs. Based on these executions, we exemplify how one can flexibly assess different aspects of a process (e.g., control flow, performance, or conformance). Using statistical tests, we eventually detect differences between subprocesses with respect to a selected aspect. Despite the approach is mostly agnostic to the decomposition applied, we present a decomposition strategy that we deem particularly suitable for PC. For this purpose, we consider the ned Process Structure Tree of a Petri net and propose a novel preprocessing approach to improve the final decomposition. We implemented the approach in ProM and evaluate it in a real-life case study.