Process mining: Using CPN tools to create test logs for mining algorithms

Process mining aims at automatically generating process models from event logs. The main idea is to use the discovered models as an objective start point to deploy systems that support the execution of business processes (for instance, workflow management systems) or as a feedback mechanism to check if the prescribed models fit the executed ones. When developing an algorithm to do process mining, one needs some logs to test it. Using real-life logs seems to be the natural choice. However, the real-life event logs usually contain imperfections that can hinder the tuning of the mining algorithm. For instance, real-life logs can be incomplete and/or contain noise. Thus, a more common approach is to first test the accuracy of new process mining algorithms in logs created via simulation. This allows the researcher to have more control about the properties of the event log and to fine tune his/her mining algorithm. Besides, having the original model (the simulated one) may also be a useful aid to assess the quality of the mining algorithm. In our research group, we work with the ProM framework [6] mining tool which receives as input an XML event log. This XML format is also supported by process mining tools of other research groups [4]. This paper shows how to extend CP-nets to generate XML event logs that can be mined by process mining tools supporting this format. This way we benefit from the simulation capabilities of CPN Tools and, therefore, we avoid reinventing the wheel. The extension we made consisted of implementing (i) some ML functions that can be used to annotate the CP-net, and (ii) a ProMimport -framework [2] plug-in that bundles up the files (generated by the CP-net simulation) into a single XML file that is ready to be mined.