Top-Down Process Mining from Multi-Source Running Logs Based on Refinement of Petri Nets

Abstract

Today's information systems of enterprises are incredibly complex and typically composed of a large number of participants. Running logs are a valuable source of information about the actual execution of the distributed information systems. In this paper, a top-down process mining approach is proposed to construct the structural model for a complex workflow from its multi-source and heterogeneous logs collected from its distributed environment. The discovered top-level process model is represented by an extended Petri net with abstract transitions while the obtained bottom-level process models are represented using classical Petri nets. The Petri net refinement operation is used to integrate these models (both top-level and bottom-level ones) to an integrated one for the whole complex workflow. A multi-modal transportation business process is used as a typical case to display the proposed approach. By evaluating the discovered process model in terms of different quality metrics, we argue that the proposed approach is readily applicable for real-life business scenario.