Schuster, DanielDanielSchusterFöcking, NiklasNiklasFöckingZelst, Sebastiaan vanSebastiaan vanZelstAalst, Wil van derWil van derAalst2023-11-292023-11-292023https://publica.fraunhofer.de/handle/publica/45736710.1007/978-3-031-41620-0_42-s2.0-851722859982-s2.0-85172184678Process discovery learns process models from event data and is a crucial discipline within process mining. Most existing approaches are fully automated, i.e., event data is provided, and a process model is returned. Thus, process analysts cannot interact and intervene besides parameter settings. In contrast, Incremental Process Discovery (IPD) enables users to actively participate in the discovery phase by gradually selecting process behavior to be incorporated into a process model. Further, most discovery approaches assume process executions, also termed traces, recorded in event data to be complete—complete traces span the actual process from start to completion. Incomplete traces are usually removed in the event data preparation as most discovery algorithms cannot handle them respectively treat them simply as full traces. This paper presents a novel IPD approach that can incorporate process behavior recorded in trace fragments, thus supporting incomplete data. Our experiments show promising results indicating that using trace fragments within IPD leads to high-quality process models.enAlignmentsProcess discoveryProcess miningconference paper