RG 5495: SOURCED – Process Mining on Distributed Event Sources

The discipline concerned with automatic process analysis techniques based on event data of complex systems is called process mining. Classical process mining has by and large assumed that event data is processed in a single, central data file on a device with sufficient computing power. In plenty of use cases event data originates from distributed, sensor-based systems that do not satisfy the assumptions of (classical) process mining in the general case. Here, events can be any kind of observations (e.g., a sensor value changed), no matter if explicitly linked to a specific activities or cases. Events occur as unbounded streams of sensed event data; they are subject to noise, inaccurate measurements and ambiguous information; and they bear the potential of correlation with background knowledge and corresponding threats to privacy by means of re-identification. As of today, the application of process mining to distributed scenarios suffers from technical and conceptual research challenges spanning the three dimensions (1) Infrastructure-awareness: The distribution and physical properties of sensor-based systems imposes specific research challenges for efficient event data processing. (2) Data-awareness: The granularity and quality characteristics of sensor data imposes specific research challenges for meaningful and privacy-sensitive event data abstraction. (3) User-awareness: The detail and complexity of sensor data imposes specific research challenges for traceable presentation and representation of distributed process mining results. The proposed research unit SOURCED – Process Mining on Distributed Event Sources will develop the methodological foundations for novel process mining techniques for distributed event data. An overarching challenge of sources process mining is the fact that the infrastructure, data, and user view can hardly be fully separated. This gives rise to benefits of collaboration. The data analysis will be enforced in the light of the data’s utility, preserving the accuracy and completeness of process mining results as much as possible by embedding the uncertainty of data in the entire process mining pipeline, making data quality issues explicit in the analysis and protecting the data through provable guarantees. The analysis focus on isolated cases will be extended and replaced by multi-dimensional event networks that can be interactively explored by user along different dimensions and corresponding visual representations. The research unit brings together expertise from the fields of process management, data and software engineering, distributed systems, and privacy mechanisms. The seven projects in SOURCED address scientific challenges at their intersections and provide a substantial research impact by developing the foundation for the next generation of process mining techniques.