Browse Teach Curate

Efficient Incremental Local Outlier Factor for Streaming Anomaly Detection

The EILOF algorithm scores each newly arriving streaming point with the Local Outlier Factor while leaving previously computed scores unchanged, avoiding the expensive neighbor-cascade recomputation of incremental LOF. The controlled inexactness is tolerable on noisy data, and experiments show lower cost and higher detection accuracy than incremental LOF as the stream volume grows.

Rosie Text

An Efficient Outlier Detection Algorithm for Data Streaming Rui Hu, Luc (Zhilu) Chen, Yiwei Wang

Proposes the Efficient Incremental Local Outlier Factor (EILOF) algorithm for online anomaly detection in data streams. The Local Outlier Factor (LOF) scores a point's outlierness by comparing its lo…