Efficient Algorithms for Finding Differences between Process Models

Abstract

Information systems from various domains record their behavior in a form of event logs. These event logs can be further analyzed and formal process models describing hidden processes can be discovered. In order to relate real and expected process behavior, discovered (constructed from event logs) and reference (manually created by analysts) process models can be compared. The result of comparison should clearly present commonalities and differences between these models. Since most process models are represented by graph-based languages, a graph comparison technique can be applied. It is worth known that graph comparison techniques are computationally expensive. In this paper, we adapt different heuristic graph comparison algorithms to compare BPMN (Business Process Model and Notation) models. These algorithms are implemented and tested on large BPMN models discovered from event logs. We show that some of the heuristic algorithms allow to find nearly optimal solutions in a reasonable amount of time.