Mining the Log-Tree of Process Traces: Current Approach and Future Perspectives

Logs recording the traces of execution of previous process instances can be exploited for different process management tasks, such as prediction and recommendation in operational support. Efficient retrieval of past traces can be very important to achieve such tasks, while building a model of the process from the log can support problem/anomaly detection and, more generally, process analysis. Trace retrieval is gaining attention in the Case Based Reasoning research community, but so far it has been faced in a completely separate way from the construction of a process model from the log, instead, we propose an approach aiming at integrating these two goals. The core notion of our proposal is the log-tree, which constitutes a "bridge" between the notions of (log) index and process model. We propose a mining algorithm to build it, and an algorithm exploiting it for trace retrieval. Future extensions of our initial contribution are also widely discussed.