Discovering Process Models from Empirical Data

Effective information systems require the existence of explicit process models; a completely specified process design needs to be developed in order to enact a given business process. This development is time consuming and often subjective and incomplete. We propose a method that discovers the process model from process logs where process events are recorded as they have been executed over time. We induce a rule-set that predict causal, exclusive, and parallel relations between process events. The rule-set is induced from simulated process log data that are generated by varying process characteristics (e.g. noise, log size). Tests reveal that the induced rule-set has a high performance on new data. Knowing the causal, exclusive and parallel relations we can build the process model expressed in the Petri net formalism. We also evaluate the results using a real-world case study.