General aim and scope
Temporally annotated event records tend to be accumulated in a continuous and massive way in the logs of disparate kinds of systems, such as social media systems, Business Process Management (BPM) and Industrial systems, human activities’ tracking systems, to cite a few.
In principle, if analyzed suitably, such log data could offer precious information on the behavior of the processes/systems that originated them, as well as to help monitor and improve the quality of these processes/systems, possibly assisting the involved people in their activities (e.g., through suitable run-time prediction/recommendation mechanisms).
However, meeting this objective is often jeopardized by several challenging issues, which include primarily:
- the dynamical and complex nature of log data (which often have a high-dimensional, temporal, and possibly non-stationary nature) and of the systems/processes that generated them;
- the lack of semantics in the log events, which are not easy to interpret in terms of relevant concepts (e.g., processes, process activities, …) and to relate to existing (high-level) models and domain knowledge;
- the uncertainty and incompleteness of the log data (due, e.g., to missing/noisy event records, as well as to the difficulty to obtain a representative sample of all possible process/system behaviors).
This special session, affiliated to the IEEE Task Force on Process Mining, is meant to offer a platform for sharing and publishing innovative research related to the problems of interpreting and analyzing complex event data (like those mentioned above), and of ultimately supporting the monitoring, analysis and improvement of the processes/systems that generated these data. The session is primarily interested in solution approaches relying on the discovery and refinement of behavioral models, or on the combination of high-level background knowledge with suitable data abstraction/interpretation techniques.