Process mining: Developing tools for discovering case-handling processes (EIT.6447)

Case-handling systems allow for more flexibility than traditional workflow management systems. The increased flexibility allows users and organizations to deviate from some predefined process, and addresses the problem that IT-support for business processes is often too restrictive. However, if the system is not strictly enforcing a predefined processes, it becomes interesting to discover what the actual process is. The idea of process mining is to discover the real process by extracting knowledge from event logs. The following problem will be tackled in this project: How to extract structured process descriptions from a case-handling system in the presence of concurrency and noise while focusing on all perspectives (i.e., control-flow, data, applications, workers, organization, etc.)? The scientific challenge is manifold. First of all, case-handling systems allow for more flexibility and as a result there is more variability and it becomes more difficult the discover the underlying process. Second, it is not known how to incorporate information from other perspectives for process mining (for example the relation between data and certain routes in the process). Third, real-life processes exhibit noise (e.g., incorrectly logged events) and concurrency (e.g., two things happing at the same time) thus complicating matters. The project should result in new algorithms and tools that address these problems. Starting point is the ? algorithm developed by the research team. Based on an event log this algorithm automatically constructs a Petri net describing the underlying process and it properties (e.g., bottlenecks). However, the ? algorithm abstracts from the other perspectives, has problems dealing with noise, and it not tailored towards case-handling processes/systems. The project will use a specific case-handling system as an experimental platform. However, many other IT products including Enterprise Resource Planning (ERP) systems such as SAP, allow users to deviate from the pre-specified process. Also for these systems it is challenging to find out how people really work and to detect differences between the actual process and the prescriptive/descriptive process model. Therefore, we anticipate that the results will be widely applicable.

To evaluate the approach and to enable practical applications, two organizations are directly involved: Pallas Athena as a technology and expertise provider and UWV for applying and evaluating the results of this project. FLOWer is the leading case-handling system developed by Pallas Athena and used by many organizations including UWV. FLOWer will be used as a concrete casehandling system to test our mining tools and algorithms on. Within UWV many processes are supported by FLOWer. These processes will be analyzed using our mining tools. On the one hand this analysis will be used to evaluate our mining algorithms and software. On the other hand this analysis will provide insight into the way people really work and generate concrete suggestions for process improvement. The results of the project will be highly relevant for other organizations developing or using "process-aware information systems". Case-handling systems are used within various organizations, e.g., payment institutions, banks, insurance companies, government bodies, telecommunications, health care, police, courts of law, etc. In these organizations, it is important to get grip on things and understand how people/departments really work. The results will also be interesting for tool developers working on case handling, workflow management, customer relationship management, and ERP software.

Er zijn 4 bedrijven bij dit project betrokken.