Analytical Fraud Detection

The goal of this project is to survey data mining methods to detect traces of suspicious acts in financial institutes. More precisely, we try to detect fraud in collaboration with a bank.

The relational nature of transactions justifies the use of relational data mining methods.

As positive training examples are not at hand, a system is currently built that allows fraud experts to visualize transaction data in various ways and to find and annotate suspicious transaction patterns using their implicit and longtime know-how.

Those transaction patterns will provide a basis for research. The most promising approaches will be built into the system and search for “hot spots”, which are then delivered to human experts for closer investigations.

Current Project Participants