Uncovering Object-Centric Data in Classical Event Logs for the Automated Transformation from XES to OCEL

Abstract

Object-centric event logs have recently been introduced as a means to capture event data of processes that handle multiple concurrent object types, with potentially complex interrelations. Such logs allow process mining techniques to handle multi-object processes in an appropriate manner. However, event data is often not yet available in this new format, but is rather captured in the form of classical, “flat” event logs. This flat representation obscures the true interrelations that exist between different objects and associated events, causing issues such as the well-known convergence and divergence of event data. This situation calls for support to transform classical event logs into object-centric counterparts. Such a transformation is far from straightforward, though, given that the information required for object-centric logs, such as explicitly indicated object types, identifiers, and properties, is not readily available in flat logs. In this paper, we propose an approach that automatically uncovers object-related information in flat event data and uses this information to transform the flat data into an object-centric event log according to the OCEL format. We achieve this by combining the semantic analysis of textual attributes with data profiling and control-flow-based relation extraction techniques. We demonstrate our approach’s efficacy through evaluation experiments and highlight its usefulness by applying it to real-life event logs in order to mitigate the quality issues caused by their flat representation.