Case Study: Capital One detects mobile app anomalies in real time with H2O.ai

Capital One catches 20,000-login spikes 4 minutes early with H2O.ai

Capital One, a major U.S. bank, needed to detect anomalies in its mobile app login traffic in real time. Their operations team struggled with traditional fixed thresholds, which could not adapt to normal volume swings and seasonal patterns. They turned to H2O.ai and used H2O Sparkling Water to build a scalable machine learning solution for this challenge.

H2O.ai's solution used a gradient boosting model (GBM) deployed via H2O Sparkling Water, which was integrated into a pipeline on AWS. This provided a live forecast of normal traffic and flagged significant deviations. The result was a measurable improvement in incident response; the system caught a spike of over 20,000 unexpected logins and triggered an alert four minutes earlier than any other alarm.

Capital One