RG 5495/1: VisualMine: Visual Analytics for Process Mining on Distributed Event Sources (SP VisualMine)

Process mining has implicitly been driven by the idea that the processes under investigation are, at least to a certain extent, structured and partially sequential. Already classical process mining in a centralized setting has struggled with these assumptions. One of the key challenges continues to be the complexity and variety of underlying processes, which often leads to spaghetti models being produced by process mining techniques. Indeed, process mining research has largely restricted itself to visual representations such as Directly-Follows Graphs and Petri nets. There are few examples of works that integrate other visual representations..

Also the process mining manifesto already emphasizes the problems of such a narrow spectrum of representations when discussing Challenge C5: Representational Bias.
The distributed setting will impose even stronger challenges in this regard and will increase problems with representational bias. The overall observation of SOURCED is that these challenges relate at least to infrastructure-awareness, data-awareness, and user-awareness. Regarding the latter, we specifically observe novel challenges related to effective visual representations due to the sheer size and hierarchical granularity of the event log data being produced. The distributed setting also bears the chance to describe causal relationships more explicitly since these are related to observable interactions between separate devices. The key research problem in this context is how can process mining on distributed event sources appropriately developed such that the respective visual representations meet the information needs of the tasks of the process analyst as a user.

With VisualMine, we contribute to the seminal work on process mining by designing and evaluating representations, algorithms, and techniques for process mining on distributed event sources.