Trouble reporting is a substantial component in any technical product's maintenance workflow. In this project, we investigated a set of methods for streamlining this workflow, using both software solutions and machine learning. The aim was to find a way of grouping trouble reports for easier analysis and other potential usecases down the line. This project was conducted in the context of telecom infrastructure. Unstructured data that is produced both by humans and machines was transformed into embeddings representations, usingmultiple Bert based language models, of which one was domain specialised. The embeddings were clustered using multiple clustering techniques, and finally labeled using machine learning. Furthermore, we compared the use of our Bert models, with the use of the classical TF-IDF representation, with the aim of creating a baseline for the performance of these models . Trials showed that the best way of representing the trouble report depended on its content. TF-IDF had benefits when the keywords were few, exclusive to the group and carried a lot of relevance. However, when the keywords had many synonyms or were counter-productive to look at, the language model showed better results. The sentence model S-Bert was almost always superior to the other Bert-based language models, even the domain specialised one.

Large software organizations handle many customer support issues every day in the form of bug reports, feature requests, and general misunderstandings as submitted by customers. Strategies to gather, analyze, and negotiate requirements are comple- mented by efforts to manage customer input after products have been deployed. For the latter, support tickets are key in allowing customers to submit their issues, bug re- ports, and feature requests. Whenever insufficient attention is given to support issues, there is a chance customers will escalate their issues, and escalation to management is time-consuming and expensive, especially for large organizations managing hundreds of customers and thousands of support tickets. This thesis provides a step towards simplifying the job for support analysts and managers, particularly in predicting the risk of escalating support tickets. In a field study at our large industrial partner, IBM, a design science methodology was employed to characterize the support process and data available to IBM analysts in managing escalations. Through iterative cycles of design and evaluation, support analysts’ expert knowledge about their customers was translated into features of a support ticket model to be implemented into a Ma- chine Learning model to predict support ticket escalations. The Machine Learning model was trained and evaluated on over 2.5 million support tickets and 10,000 escalations, obtaining a recall of 79.9% and an 80.8% reduction in the workload for support analysts looking to identify support tickets at risk of escalation. Further on- site evaluations were conducted through a tool developed to implement the Machine Learning techniques in industry, deployed during weekly support-ticket-management meetings. The features developed in the Support Ticket Model are designed to serve as a starting place for organizations interested in implementing the model to predict support ticket escalations, and for future researchers to build on to advance research in Escalation Prediction. / Graduate