View article

More and more information about processes is recorded by information systems in the form of so-called “event logs”. Despite the omnipresence and richness of these event logs, most software vendors have been focusing on relatively simple questions under the assumption that the process is fixed and known, eg, the calculation of simple performance metrics like utilization and flow time. However, in many domains processes are evolving and people, typically, have an oversimplified and incorrect view on the actual business processes. Therefore, process mining techniques attempt to extract non-trivial and useful information from event logs. One element of process mining is control-flow discovery, ie, automatically constructing a process model (eg, a Petri net) describing the causal dependencies between activities. Many control-flow discovery techniques have been proposed in literature. Unfortunately, these techniques have problems when discovering processes with complicated dependencies. This paper proposes a new two-step approach. First, a transition system is constructed which can be modified to avoid over-fitting. Then, using the “theory of regions”, the model is synthesized. The approach has been implemented in the context of ProM and uses Petrify for synthesis. This paper demonstrates that this two-step approach overcomes many of the limitations of traditional approaches.