Machine learning and artificial intelligence (AI) have a long track record of successful use in payment fraud. However, the hype can get overwhelming and it can get difficult for many to distinguish myth from reality. Machine learning can clearly distinguish legitimate and fraudulent behaviours if done properly, while adapting over time to new, previously unseen fraud tactics.
A lot of times, instead of relying on the importance of domain knowledge, the providers of fraud analytics rely on generic behaviour models based on relatively few cases in the model development process that must learn to identify patterns of fraud slowly over time. Consider this example: A 42-year-old woman from Sacramento, CA is a frequent domestic traveller and is attempting to withdraw the equivalent of $300 US from an ATM in Seoul. In that case, your fraud system has less than a second to make a risk determination.
This behaviour might be anomalous and it is relatively easy to determine. But is it indicative of fraud? That’s a tougher question that only specialized fraud analytics which is honed on huge quantities of data can accurately assess.
Specialized fraud analytics must be used to assess the “tough” questions in order to maintain a positive consumer experience. This is where advanced profiling, fraud-specific predictive characteristics, and adaptive capabilities separate themselves from generic behaviour analytics.
Generic behaviour models are not sufficient for cross-channel, enterprise fraud solutions in a world of real-time payment processing and rapidly changing consumer preferences. After all, when and how someone chooses to transact is not as predictable as his or her likelihood to cancel a fitness club membership.