Algorithms for detecting anomalies in River

In this chapter, you will again use River for online machine learning algorithms. There are other libraries out there, but River is a very promising candidate for being the go-to Python package for online learning (except for reinforcement learning).

You will see two of the online machine learning algorithms for anomaly detection that River currently (version 0.9.0) contains, as follows:

• OneClassSVM: An online adaptation of the offline version of One-Class SVM
• HalfSpaceTrees: An online adaptation of Isolation Forests

You will also see how to work with the constant thresholder and the quantile thresholder.

The use of thresholders in River anomaly detection

Let's first look at the use of thresholders, as they will be wrapped around the actual anomaly detection algorithms.

Anomaly detection algorithms will generally return a score between 0 and 1 to indicate to the model to what extent the observation is...