Discovering workflow nets of concurrent iterative processes

Abstract

Abstract A novel and efficient method for discovering concurrent workflow processes is presented. It allows building a suitable workflow net (WFN) from a large event log $$\lambda $$ λ , which represents the behaviour of complex iterative processes involving concurrency. First, the t -invariants are determined from $$\lambda $$ λ ; this allows computing the causal and concurrent relations between the events and the implicit causal relations between events that do not appear consecutively in $$\lambda $$ λ . Then a 1-bounded WFN is built, which could be eventually adjusted if its t -invariants do not match with those computed from $$\lambda $$ λ . The discovered model allows firing all the traces in $$\lambda $$ λ . The procedures derived from the method are polynomial time on $$|\lambda |$$ | λ | ; they have been implemented and tested on artificial logs.