Detecting different types of workarounds in IT systems with clustering algorithms

Abstract

This research shows how and to what extent data clustering algorithms can be used to detect different types of workarounds in structured data of IT systems. Finding them can reduce the risk of lacking data quality and offers opportunities for system improvements. Existing workaround studies place more emphasis on process mining and techniques such as observations, questionnaires and interviews to find them. The Resilience Mining Masters Circle provides a complete research for detecting workarounds in IT systems with data mining techniques. This study focuses in particular on detecting different types of workarounds with clustering algorithms using the CRISP-DM approach.
Three experiments have been set up within an artificial data set to detect expressions of the workaround types fictious entity, overcome inadequate IT functionality and misuse of a (text) field with the clustering algorithms k-Means (fast), k-Medoids, Agglomerative Clustering and DBSCAN.
The experiments show that multiple clustering algorithms are suitable for detecting the workaround types fictious entity and misuse of a (text) field. The type of overcome inadequate IT functionality was not found in this study. Because this was an artificial data set, it is advisable to repeat this research on a real-life data set and with other expressions of the workaround types examined to solidify this conclusion.

Date of Award	5 Jun 2020
Original language	English
Supervisor	Lloyd Rutledge (Examiner) & Guy Janssens (Co-assessor)