This thesis explores domain adaptation methods to improve machine learning models in a networking environment where the domain of the data changes after initial data gathering. The applied methods make use of Generative Adversarial Nets and aim to create a model to predict service performance on a client machine using resource utilisation data collected from a server cluster that provides the service. However, since the models were originally designed for image classification, they were changed to work on the problem at hand, because it is a regression task with tabular data. The report explores the theory behind these methodologies and presents the implementations and their results in this project. Unfortunately, the chosen methods did not provide an improvement compared to a regular model, therefore the report also examines the possible causes for this, and provides informed reasons.