PAIS Lab seminar: Method for synthesis of well-structured process models for transition systems based on hierarchy of events

At the session of our seminar, Evgenii Elokhov (Master's student at the HSE CS Faculty) delivered a report "Method for synthesis of well-structured process models for transition systems based on hierarchy of events"

Abstract:
It is well known that Petri nets are widely used for modeling systems. Transitions in Petri nets can be seen as events occurring in the system. There are many algorithms for Petri nets discovery: for example, an algorithm which generates a model of the process from the corresponding transition system using the theory of regions. Transition systems can be "discovered" from event logs, which is a set of traces, each of which describing a sequence of events. An existing algorithms of Petri nets discovery, based on regions, is capable of discovering precise process models, which can replay all the traces from the event logs, but at the same time, these models may not have a clear structure and not be generalized.

In tis work, an algorithm, based on regions and events hierarcy, for discovering well-structured process models is proposed. This subject is important for the following reason: events hierarchies allow for removal of excessive places, which is not possible when using classical algorithms for synthesis of Petri nets. Any process model generated by the proposed algorithm should be able to replay all the traces from the event. Furthermore, the structure of the obtained model should be easier compared with the model, resulting from the application of the classical algorithm. However, the approach based on the use of a hierarchy of events adds a new behavior pattern, making it less accurate. The proposed algorithm has been implemented as a Java plug-in for the ProM framework.