Automatic Detection of Errors and Anomalies in Data

The bachelor thesis focuses on the automated detection of data issues, with an emphasis on the use of modern technologies to ensure data quality. The aim of the thesis is to present automation possibilities for identifying common data problems such as missing or invalid values, outliers and anomalies, schema drift, and data drift. The theoretical part defines key concepts, dimensions of data quality, and tasks related to its management, with special attention given to the concept of augmented data quality, which integrates traditional approaches with advanced tools such as machine learning, natural language processing (NLP), and large language models (LLMs). In the practical part, various methods for detecting data issues in a real dataset are proposed and tested. The approaches include rule-based systems, machine learning algorithms, automated profiling tools, and the use of LLMs. The results show that automated detection is both technically feasible and practically beneficial, especially when combining multiple methods. Rule-based approaches provide transparent and easily interpretable outputs, while large language models offer greater flexibility and the ability to capture more complex patterns in data. However, their use requires careful validation, as they may produce inaccurate or redundant results without contextual adaptation. The thesis also identifies directions for further development in the field of data quality automation, such as systematic testing of LLM outputs on annotated data, exploring the impact of input contextualization, or integrating detection tools into broader data ecosystems. The findings of this thesis can serve as a starting point for further research and practical deployment of tools for effective data management.