Teradata announced today that Danske Bank has worked with Think Big Analytics, a Teradata company, to create and launch a state-of-art, AI-driven fraud detection platform that is already expected to meet 100 percent ROI in its first year of production.
The engine uses machine leaning to analyz tens of thousands of latent features, scoring millions of online banking transactions in real-time to provide actionable insight regarding true, and false, fraudulent activity. By significantly reducing the cost of investigating false-positives, Danske Bank increases its overall efficiency and is now poised for substantial savings.
Danske Bank’s original fraud detection system was largely based on handcrafted rules that had been proactively applied by the business over time. With record numbers of false positives – at times reaching 99.5 percent of all transactions – the costs and time associated with investigation had become significant, with the bank’s large fraud detection team feeling overworked, yet not effectively utilized.
Teradata’s Think Big Analytics team began working with Danske Bank in autumn 2016, to augment their advanced analytics team with specialist knowledge about how to utilize data to bring greater benefits to the wider business. The joint team began with building a framework within the bank’s existing infrastructure and then created advanced machine learning models to detect fraud within millions of transactions per year, and in peak times, many hundreds of thousands per minute. To ensure transparency and encourage trust, the engine includes an interpretation layer on top of the machine learning models, providing explanations and interpretation of blocking activity.
From a modeling point of view, fraud cases are still very rare, with around one fraud case in every 100,000. The team has managed to take the false positives from the models and reduce them by 50 percent. At the same time, they are able to catch more fraud – actually upping the detection rate by around 60. Danske Bank’s anti-fraud program is the first to put machine learning techniques into production while simultaneously developing deep learning models to test the techniques.