University of Twente Student Theses
A Case Study of AI performance enhancement in Incident Management processes with low data-quality
Stomp, D.D (2024) A Case Study of AI performance enhancement in Incident Management processes with low data-quality.
|
PDF
482kB |
| Abstract: | Artificial Intelligence has created new possibilities to increase operational efficiency in many fields, including the field of IT Service Management. Specifically in the resolution process, which focuses on the resolution of issue tickets. These tickets are essentially a bundle of correspondence between the IT agent and the problem holder. These tickets are categorized manually to generate data for continuous improvement and for further escalation down the escalation paths of the support team. AI is already able to classify these tickets based on the initial issue description, replacing the need for manual classification, and therefore increasing operational efficiency. However, some businesses suffer from low quality data sets, which, proven in prior research, will negatively affect the classification capabilities of AI. It was also proven that addressing the issue of low data quality in a data set through extensive data analysis will improve the performance of the corresponding AI models. However, for small to medium-sized businesses without the in-house capabilities for such methods, easier methodologies had to be explored. In this case study, such a low-quality data set was provided by an IT service provider in the nautical tourism sector and enhanced in multiple ways to measure their effect on AI classification performance. For evaluation, AI models had to be created, tested, and trained on the native and enhanced versions of the data set. The results of this research show that minor gains in performance can be achieved through systematic changes in the data set, like a better separation of categories based on their semantic meaning. The larger gains were achieved by removing the lowest quality entries from the data set. This was done, alternatively to extensive data analysis, by having an expert from the support team look for the common indicators of low-quality entries. |
| Item Type: | Essay (Bachelor) |
| Faculty: | BMS: Behavioural, Management and Social Sciences |
| Subject: | 85 business administration, organizational science |
| Programme: | International Business Administration BSc (50952) |
| Link to this item: | https://purl.utwente.nl/essays/100164 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page