Thursday, 7 June 2012, 14:00
Cybernetica Bldg (Akadeemia tee 21), room B101
Slides from the talk [pdf]
Abstract: The purpose of process mining is to construct or to reconstruct from an event log a business process model that can generate this event log. The game is to dig out of event logs sufficient informations on the structure of their generating model. As a technique for model discovery, process mining has some connections with machine learning.
Process mining may be used for the purpose of modelling. For instance, after collecting over a long period of time information on the health history of many patients, including the diagnosis and treatment steps, one may want to extract from this record an accurate model of the workflow system of an hospital. Reverse engineering, which consists of reconstructing from representative use cases an existing but partially unknown system, is another activity of model discovery that can be achieved by process mining. Process mining can also be used for conformance checking or enhancement of business process models. For instance, process-aware systems record run-time informations used to detect discrepancies between expected and actual behaviours and to refactor these systems.
The process mining algorithm α was introduced by van der Aalst et al for the discovery of workflow nets from event logs. This algorithm was presented in the context of structured workflow nets even though it was recognized that more workflow nets should be reconstructible. In this presentation we assess the α algorithm and provide a characterization of the class of workflow nets reconstructible by α. We also provide a comparison of α with an alternative process mining algorithm using the region-based synthesis of elementary net systems.
This talk is given in the frame of the 4th French month of science in Estonia.