What is Anomaly Detection in Machine Learning?

Anomaly Detection refers to identification of rare occurances in the data and may be achieved through supervised, unsupervised or semisupervised methods.

Anomaly detection via supervised methods is similar to the classification with the exception of unbalanced nature on inputs where anomalies are very rarely present.

Anomaly detection via unsupervised methods may use either clustering, associative rule learning, or mapping data to lower dimensions and label samples differing widely from rest of data as anomalies.

Finally a wide ranging set of approaches in semi-supervised learning may be used to infer information from the small set of labeled anomalies on the rest of data.