Project abstract

The goal of this project is to consistently detect implausible, atypical or unusual constellations in the landscape of financial institutions. Specifically, in Austria more than 500 financial institutions report data on their entire credit portfolio to the regulator on a monthly basis. The regulator faces the challenge to transform extensive amounts of data into an (human) accessible representation of the underlying business models. Building on a pilot project in cooperation with the OeNB, we aim to develop a set of unsupervised learning methods which address this challenge in a stable and time-consistent manner. To achieve this we rely on novel tools and stability results from the emerging theories of entropic and adapted transport.