WiSe26-27: Software Engineering (B.Eng.)

Data Science: from a production database to a conclusion you can defend

Most of the work in data science, and most of the risk, comes before a model is fitted. This module covers that part: where data comes from, what is wrong with it, what may legitimately be concluded from it, and how to say so in a way that holds up to questions. All work uses one real production database (AdventureWorks Cycling in MySQL), not a prepared teaching example. No models are trained; the module ends where a machine learning course begins.

Topics: tidy data and joins; turning a query into a repeatable extraction; formats, metadata and personal data; data quality from profiling to a documented cleaning log; time series and graphs; features, similarity and distance; PCA, t-SNE and clustering; exploratory statistics and visualisation; dashboards and ML-readiness. You can work in Python or in KNIME, and either one is enough.

The lecture is held in English. Slides and exercise sheets are available in English and German. You should know SQL and the relational model, and have basic programming skills. Statistics is not required.

Assessment is a group portfolio. Each group works on its own slice of the database and makes three submissions during the semester, followed by an oral exam. Start with the Onboarding tile.