Identifying Petri Nets with Silent Transitions by Event Traces Classification

摘要

A method for discovering workflow nets (WFN) with silent transitions from a log of event tracesλis presented. It operates in two stages; in the first one,λis partitioned into three classes of traces: normal traces, short abnormal traces, and long abnormal traces. In the second stage, the normal traces are processed to build a WFN that may contain transitions of typeinitializeandfinalize;afterwards, this net is refined by adding the transitions of typeskipandredo,which are determined from the short abnormal traces and the long abnormal traces respectively. Implementation and tests are presented.