Discovering Unbounded Synchronization Conditions in Artifact-Centric Process Models

Automated process discovery methods aim at extracting business process models from execution logs of information systems. Existing methods in this space are designed to discover synchronization conditions over a set of events that is fixed in number, such as for example discovering that a task should wait for two other tasks to complete. However, they fail to discover synchronization conditions over a variablesized set of events such as for example that a purchasing decision is made only if at least three out of an a priori undetermined set of quotes have been received. Such synchronization conditions arise in particular in the context of artifact-centric processes, which consist of collections of interacting artifacts, each with its own life cycle. In such processes, an artifact may reach a state in its life cycle where it has to wait for a variable-sized set of artifacts to reach certain states before proceeding. In this paper, we propose a method to automatically discover such synchronization conditions from event logs. The proposed method has been validated over actual event logs of a research grant assessment process.