AI case study: fraud detection gets stronger at Visa

Aug 17, 2021 | Artificial intelligence

To tackle fraud, Visa used an AI approach to get a 20-30% lift in performance – and in some cases even 100%. It applied advanced AI techniques versus more traditional techniques like gradient boosted trees. Collectively this prevented $25bn in fraud.

Visa’s ‘Advanced Authorization’ platform scores every transaction that goes across their network and rates each one based on the likelihood that it’s fraudulent.

By improving the separation between good and bad transactions, the system allows more transactions to be approved more quickly. Both AI and machine learning tools were introduced in a layered approach in systems outside its main transaction processing network – which avoided increasing latency. Here’s more about how it worked…

Visa went on to use AI and proprietary data to help financial institutions predict credit application fraud. The technology generates an identity fraud score in near real-time, which prevents fraud loss at the moment of a credit or loan application. See the details here…

All topics

Previous editions

Get email edition

Share This