Mining Extended Event Structures from Logs

Process Mining is a set of methods aimed at the analysis of historical recorded data, or event logs, and extracting the knowledge about the corresponding business processes [20]. Process Mining allows analysts to understand how a process is executed as well as to find out deviations with respect to the expected behaviour. Process Mining has gained considerable practical adoption in recent years, as confirmed by increasing number of case studies and software tools [5]. Nevertheless, the Process Mining field has evolved in an ad-hoc way, i.e. without a foundational theory that would allow methods and results developed for one problem to be re-used when handling other similar cases [8]. In [8] authors propose to use well-known model of concurrency, namely Event Structures, as a unifying theoretical basis for Process Mining techniques. In this work we propose to extend the notion of Prime Event Structure with two additional relations - concurrency and recently introduced by Stefan Haar et. al. in [16] reveals relations. We investigate the problem of mining Extended Event Structures from logs and using it in conformance checking (i.e. checking if processes are executed in accordance with the specified requirements) and related tasks. Both new and existing techniques are evaluated. The proposed approaches and concepts are implemented as a proof-of-concept tool. As assessments presented in Chapter 4 show, proposed method for conformance checking is able to provide additional information on the process insights, specifically it: - gives reasonable results that correlate with corresponding results produced by traditional approaches; - is more sensitive in terms of relations between events that could indicate possible insights for the violations in a process executions; - for some kind of processes, gives more relevant results that the traditional methods as it shown in the running example.