Comparing Complex Business Process Models

Business process models extracted from real event data via process mining are often visually complex and hard to interpret, cannot easily display all of the relevant process data in one view, and are of unknown quality. It is therefore difficult to present a meaningful comparison of two mined process models, for example to confirm business changes have had a significant effect. We propose a framework for comparing process models, which integrates metrics, graph partitioning, notions of statistical significance, and visualisation techniques. This framework enables well-founded and intuitive methods for understanding and exploring differences between processes. Keywords—Business process mining, graph partitioning, visualisation, probability distributions.