Discovering and navigating a collection of process models using multiple quality dimensions

Abstract

Process discovery algorithms typically aim at discovering a process model from an event log that best describes the recorded behavior. However, multiple quality dimensions can be used to evaluate a process model. In previous work we showed that there often is not one single process model that describes the observed behavior best in all quality dimensions. Therefore, we present an extension to our flexible ETM algorithm that does not result in a single best process model but in a collection of mutually non-dominating process models. This is achieved by constructing a Pareto front of process models. We show by applying our approach on a real life event log that the resulting collection of process models indeed contains several good candidates. Furthermore, by presenting a collection of process models, we show that it allows the user to investigate the different trade-offs between different quality dimensions. © Springer International Publishing Switzerland 2014.