Automatic Multi-Perspective Conformance Checking for BPMN Models

Abstract

Ahmed T. Suliman , “Automatic Conformance Checking for BPMN models', M.Sc. Thesis, M. Sc. in Electrical and Computer Engineering, Department of Electrical and Computer Engineering, Khalifa University of Science and Technology, United Arab Emirates, July, 2018. In the last decade, there has been a massive growth in the area of Business Process Management (BPM). As business processes become more complex, they generate a huge amount of data. The goal of BPM is to improve business processes' efficiency and reduce errors. At the heart of BPM is process mining, which is concerned with the analysis of process executions. A major goal of process mining is ensuring that a process is executed correctly by comparing a process model and an event log. This is known as off-line conformance checking. Previous work in conformance checking focuses on formal models such as Petri nets as opposed to visual business modeling languages used in practice, such as BPMN. Additionally, available techniques are computationally expensive and require much effort on the part of business analysts. Finally, such work tends to ignore temporal and organizational constraints. e.g. concerning resource availability. In this thesis, a multi-perspective online conformance checker for BPMN models is presented. The proposed conformance checker is able to validate process in terms of control-flow, temporal and organizational rules. This is performed through two main steps. Firstly, BPMN models are formalized through conversions to Petri nets, and Control-Flow rules are extracted from the generated Petri nets. These rules are then augmented with additional constraints specified by the BPMN models. The second step is to check the generated rules against an event log to evaluate its conformance while it is being generated. We achieve this by adapting the architecture of Snort, a popular Intrusion Detection System. The proposed framework is validated through experiments on both real and synthetic datasets. Additionally, the framework is compared to existing tools for conformance checking. The experiments show that the framework is able to detect violations accurately and efficiently.

Date of Award	Jul 2018
Original language	American English
Supervisor	Ernesto Damiani (Supervisor)