Isolation Forest

A high-performance unsupervised algorithm that identifies anomalies by isolating outliers through random recursive partitioning.

Isolation Forest (introduced by Liu, Ting, and Zhou in 2008) shifts the focus from profiling normal data to explicitly isolating anomalies. The algorithm constructs an ensemble of Isolation Trees (iTrees) where outliers are partitioned into significantly shorter paths: their scarcity and distinctiveness make them easier to separate than normal points. It maintains a linear time complexity (O(n)) and a low memory footprint: making it the industry standard for real-time applications like credit card fraud detection and network security. By avoiding expensive distance or density calculations, it remains effective even as the dimensionality of the dataset increases.

https://scikit-learn.org/stable/modules/outlier_detection.html#isolation-forest