The fight against payment card fraud resembles an arms race. Card issuers are deploying ever more sophisticated anti-fraud measures, and fraudsters are continually evolving strategies to evade those measures.
Typically, issuers rely upon neural network fraud models trained on huge historical datasets to recognize recurring fraud patterns and reduce fraud losses. However, the fraudsters’ decentralized nature and constant adaptation give them an evolutionary advantage over the issuers’ multi-month to multi-year analytic development cycles.
So what’s an issuer to do?
The answer lies in deploying adaptive analytics. When used with neural network models, these swing the advantage back to the issuers by continually adapting the fraud detection models based on the latest fraud behavior in production. This not only improves model performance, but also extends the useful lifetime of static neural network models.
We describe how adaptive analytics work in a FICO Insights white paper (which requires registration to download). In a test described in the paper, adaptive modeling techniques improved fraud account detection by nearly 20% and real-time value detection by more than 15%, at a 10:1 account false positive ratio.
