Context-aware detection of deviations in process executions

Contact

Name

Gyunam Park

Scientific Assistant

Phone

work
+49 241 80 21908

Email

E-Mail
 

MSc Thesis Project

Title: Context-aware detection of deviations in process executions

Supervisors: Gyunam Park

Examiner: Prof. Wil M.P. van der Aalst

2nd Examiner: Prof. Martin Grohe

Summary

Deviation detection in process executions aims to identify deviating behaviors that negatively affect the performance of the process. It is essential to consider the process context for relevant and effective deviation detection, because process executions occur in different contexts. Hence, several methods for context-aware deviation detection have been proposed. However, each existing work has enhanced a specific deviation detection method with a limited set of contexts such that they lack a comprehensive integration of deviation detection and contextawareness in a general way.

This work presents a framework for context-aware deviation detection to support such a comprehensive integration. The framework does not restrict the large space of existing deviation detection methods and the large space of potentially interesting contexts. To cover all deviation detection methods and contexts, we define deviation detection based on a taxonomy of deviation detection methods and context using a context ontology. To connect both without any restrictions, we develop a novel, sound, and general mechanism to integrate the deviation detection world with the process context world.

In more detail, the integration mechanism is based on a novel distinction between contexts that explain deviating behaviors and contexts that cause normal behaviors to become interesting. In context-aware deviation detection, the deviating behaviors that are explained by the former contexts become context-aware non-deviating, e.g. a deviating process execution in an exceptional context of an online-shop’s sales week is explained by the sales week leading to a context-aware non-deviating process execution, and the normal behaviors with the latter contexts become context-aware deviating, e.g. a normal package delivery process execution that is forgotten for a few days leading to an exceptional waiting time becomes context-aware deviating.

The framework is implemented as a containerized web application for the proof of concept. Furthermore, the implementation features a dedicated, supportive, web-based user interface that guides the user in navigating the many possibilities of context-aware deviation detection and enables root cause analysis. Besides, the framework is evaluated using the implementation by conducting four case studies in a simulated business process.