Beginning Anomaly Detection Using

 Isolation  Forest

An isolation forest is a collection of individual tree structures that recursively partition 

the data set. In each iteration of the process, a random feature is selected, and the data 

is split based on a randomly chosen value between the minimum and maximum of 

the chosen feature. This is repeated until the entire data set is partitioned to form an 

individual tree in the forest. Anomalies generally form much shorter paths from the 

root than normal data points since they are much more easily isolated. You can find the 

anomaly score by using a function of the data point involving the average path length.

Applying an isolation forest to an unlabeled data set in order to catch anomalies is an 

example of