Deducing Case IDs for Unlabeled Event Logs

Abstract

Event logs are invaluable sources of knowledge about the actual execution of processes. A large number of techniques to mine, check conformance and analyze performance have been developed based on logs. All these techniques require at least case ID, activity ID and the timestamp to be in the log. If one of those is missing, these techniques cannot be applied. Real life logs are rarely originating from a centrally orchestrated process execution. Thus, case ID might be missing, known as unlabeled log. This requires a manual preprocessing of the log to assign case ID to events in the log.

In this paper, we propose a new approach to deduce case ID for the unlabeled event log depending on the knowledge about the process model. We provide a set of labeled logs instead of a single labeled log with different rankings. We evaluate our prototypical implementation against similar approaches.