Anomaly detection

Anomaly Detection (Outlier Analysis) is the ML-driven process of flagging rare data points or events that significantly deviate from an established normal pattern.

Anomaly Detection identifies critical incidents by modeling a system's 'normal' behavior, then using statistical methods or Machine Learning (ML) to flag significant deviations as outliers. This is a vital, proactive capability across multiple sectors. For example, financial institutions use it for real-time credit card fraud detection, flagging transactions (e.g., a $5,000 purchase in a new country) that violate a user's spending baseline. In IT Operations, it monitors infrastructure metrics (CPU, latency) to predict equipment failure or system health issues before a total outage. The goal is simple: reduce false positives while ensuring early detection of high-impact events like a network intrusion or a manufacturing defect.

https://www.ibm.com/topics/anomaly-detection