This is actually amazon’s algorithm used to detect anomalies. It works with current streaming data as well as offline data. 

University of Michigan - Ann Arbor

Robust random cut forests also isolate outliers, however they do so through association. It associates a score to an anomaly, thereby analyzing these scores on a basis of “normal”, “low” or “high”. It works extremely well on robust data. 


Robust random cut forests 


Robust random cut forest applications

Robust random cut forest code example

http://proceedings.mlr.press/v48/guha16.html
http://proceedings.mlr.press/v48/guha16.pdf

Learn Before

Related