Context. When developing or maintaining large component-based software systems (for example, microservices, which are getting more and more popular1), software developers are usually divided into multiple developer teams working on different components. Additionally, it is not uncommon to include third-party components in one’s software. During the development and maintenance of such a system, issues might occur. Those bugs might not harm the component they arose in but might cause failures on neighbouring ones. Traditionally, this is avoided by manually annotating or linking issues across components to indicate their relationships. Problem. In large projects, this is hardly feasible due to the lack of knowledge a single developer has over the whole system as well as the other issues/components involved. Objective. This bachelor’s thesis tries to predict issue relations using machine learning (ML), which solves the generalisation of the problem. The objective hereby is to create a data set of annotated issue relations, train ML models on it, as well as measure their performance. Method. ML models have to be trained on an existing data set of annotated issues and their relations. Those samples of issues and relations are obtained from public GitHub repositories by scraping and inferring their relations. Issue relations can be categorised into five distinct categories, namely: (1) “issue A depends on issue B” (2) “issue B depends on issue A” (3) “issue A and B do not stand in a relation” (4) “issue A and B have a mutual relation” (5) “issue A is a duplicate of B” Afterwards, a baseline model will be created to test other models against it and measure their performance. Conclusion. The conclusion would be to determine if and to what extent the method presented above will lead to good results.