Anomaly detection finds outliers and novelties, defined as observations that are significantly different from the others.

Outlier detection estimators try to fit the regions where the training data is the most concentrated, ignoring the deviant observations. Novelty detection finds whether a new observation is an outlier, in which case an outlier is also called a novelty. However, novelty detection assumes no outliers in the training data.

An example use case can be analyzing all environmental or traffic monitoring sensor data to find out the anomaly that could lead to identifying dysfunctional sensors.

Spatial anomaly detection identifies observations that are geographically isolated using spatial weights with standard anomaly detection methods (see the Local Outlier Factor anomaly detection method). In addition, Spatial Clustering and Regionalization algorithms can also be used to detect outliers and analyze anomalies.