Inferring the Repetitive Behaviour from Event Logs for Process Mining Discovery

This paper addresses the problem of discovering a sound Workflow net (WFN) from event traces representing the behavior of a discrete event process. A novel and efficient method for inferring the repetitive behaviour in a workflow log is proposed. It is based on an iterative search and filtering of cycles computed in each trace; a graph of causal relations is built for each cycle, which helps to find the supports of the t-invariants of an extended WFN. The t-invariants are used for determining causal and concurrent relations between events, allowing building the WFN efficiently in a complete discovery technique.