A very powerful anomaly detection method has been proposed by Liu F T, Ting K M, and Zhou Z, in the article Isolation Forest, ICDM 2008, Eighth IEEE International Conference on Data Mining, 2008) and it's based on the general framework of ensemble learning. As this topic is very wide and mainly covered in supervised machine-learning books, we invite the reader to check one of the suggested resources if necessary. In this context, instead, we are going to describe the model without a very strong reference to all the underlying theory.
Let's start by saying that a forest is a set of independent models called decision trees. As the name suggests, more than algorithms, they are a very practical way to partition a dataset. Starting from the root, for each node, a feature and a threshold are selected and the samples are split into two...
