Anomaly Detection Guide


Def:

Anomaly Detection is the technique of identifying rare events or observations which can raise suspicions, deviate significantly from the majority of the data

Concepts:

Point & Collective

Local Anomaly:

A Local Anomaly is the observation group that is deviate from the NN(nearest Neighbourhood, it could be a point or collective), as figure above, is a local anomaly point

Global Anomaly:

The observation group that is deviate from the all observations, and


Method:

KNN Anomaly Detection Isolation Forest

notes: isolation forest has nothing to do with the random forest


Key words:


TAGS