July 2019
Beginner to intermediate
740 pages
16h 52m
English
The isolation forest algorithm is a random forest where the splits are made on randomly chosen features. A random value of that feature between its maximum and its minimum is selected to split on. Note this range is from the range of the feature at that node in the tree, not the starting data.
A single tree in the forest will look something like the following:

The average length of the path that must be traveled from the top of each tree in the forest to the leaf containing a given point is used to score a point as an outlier or inlier. The outliers have much shorter paths, since they will be one of the few on a given side ...
Read now
Unlock full access