Process-aware information systems, e.g., workflow management systems and Enterprise Resource Planning (ERP) systems are becoming more mature and, as a result, the interests of users and developers are shifting from automation of processes to process improvement. Therefore, there is a clear interest from industry for the topic of "workflow mining," or as we prefer to call it, "process mining." The reason why many scientists are interested in the topic of process mining is that it is a new and exciting area which offers a lot of scientific challenges.

The paper discusses a number of approaches. The Alpha algorithm takes, for example, a simple idea that works very well in many situations. As a result, it can be applied in different fields. Also the XML format (see also www.processmining.org) defined in this paper can be used to join efforts of many scientists.

This paper provides a survey of some of the existing approaches and highlights the challenges of process mining. The basic idea of process mining is to discover a process model on the basis of a so-called "event log." In a way process mining is a "kind of magic;" by identifying patterns in the log we are able to discovery process models. In contrast to classical data mining, we do not consider static properties but focus on the dynamics. This creates many challenges, e.g., how to detect parallelism and how to discover that things causally depend on one another. A nice example is the set of event logs in hospitals. By combining these logs we can discover typical treatment processes for specific patient groups.

In our research group we wanted to combine the expertise of Ton Weijters in the area of machine learning with my expertise in the area of workflow systems and Petri nets. This was when we discovered the topic of process mining and defined a Ph.D. project. At this point in time, three Ph.D. students in our group are working on this and one already completed her Ph.D. on the topic. Moreover, other research groups are joining our efforts by making plug-ins for the ProM (process mining) framework.