Unsupervised anomaly detection

Unsupervised anomaly detection automatically flags critical outliers in unlabeled data, identifying deviations like a 98% drop in web traffic or a zero-day security threat without requiring prior examples.

This technology is a machine learning powerhouse: it profiles normal system behavior, then isolates data points that significantly deviate from that learned baseline. Key algorithms like Isolation Forest, One-Class SVM, and Local Outlier Factor (LOF) are deployed to find the rare exceptions. We use it extensively in finance for credit card fraud detection, in IT for spotting server performance anomalies, and in manufacturing to detect equipment faults before total failure: it’s the essential tool for finding the unknown unknowns in massive, unlabeled datasets.

https://www.builtin.com/data-science/anomaly-detection-algorithms