Nowadays, data is the basis of many processes and decisions in industry and research. In the area of quality management in production, for example, data analyses can be used to determine the causes of defective products and subsequently repair the corresponding components of the products in a targeted manner.

In practice, such industrial data in particular often has specific characteristics that lead to various challenges for data analysis. If the existing data characteristics are not addressed accurately during data preparation, the direct application of analysis algorithms to the then inadequately prepared data leads to a moderate informative value. Therefore, experts in the field of data science are needed who have in-depth knowledge of the methods and algorithms involved in preparing and analyzing data. However, these experts usually lack the necessary domain knowledge, e.g., knowledge of the various products and the dependencies between various components of the respective products, so that the data can be prepared accordingly and analyzed profitably. This knowledge is difficult to define even for domain experts and therefore remains mostly unused in data analysis.

This project deals with data characteristics that often occur in industrial use cases. Therefore, it investigates how a targeted data preparation can be used to address such data characteristics. If several of these data characteristics are present in combination, purely data-driven methods are usually not able to address them sufficiently. Therefore, it will be explored how existing domain knowledge of the industry partner can be used in a targeted way to enable more meaningful analysis results. This will then be investigated and evaluated on the basis of real use cases of the industry partner.