In order to prevent fraud, the salary slips of financing recipients are now checked manually on a random basis according to a heuristic. In order to optimise internal processes, the number of salary statements to be checked manually is to be reduced. The fraud detection rate should not deteriorate in the process.
A machine learning model is being developed to estimate the probability of fraud in financing transactions. Self-disclosure data and information on the vehicle and financing product form the data basis. The optimal threshold value for triggering transactions for manual verification is determined.
The number of salary slips to be manually checked can be reduced by over 50% - and the fraud detection rate remains the same. The model is implemented in the production system.
Are you interested in your own use cases?
Our Case Studies
- Get even more detailed insights into our customer projects -
Smart cooking with the Thermomix
Data Operations at Munich Re
Data & AI Knowledge
Creating added value from data & AI together