Every year, Germany loses billions of euros in tax revenue due to cross-border VAT fraud and tax arrangements with international implications. The detection and investigation of such cases proves to be very complicated and time-consuming. On the one hand, the partially unstructured data that a tax auditor must process and analyze must be mentioned. On the other hand, the sometimes insufficient data situation is an obstacle, especially in the fight against cross-border VAT fraud, which can also be the result of organized crime.
The goal of the research cooperation between the VLBA and the State Tax Office, which was launched in December 2020, is the early detection of sales tax fraud cases and aggressive tax avoidance practices with the help of data analytics.
In this application context, data science methods can provide a way to greatly increase the size of the data sets under consideration by automating manual audit steps, and in addition, automating previously manual activities also reduces the time required. The optimizations concern the selection of suitable cases, the preparation of a test, i.e. the collection and preparation of a test, i.e. data collection and pre-processing, as well as the actual test. actual testing. While manual testing steps can be automated, new can be automated, new insights can also be gained from the data, for example, through the implementation of anomaly anomaly detection.