anomaly detection

Anomaly detection, in computer science, is the process of identifying data instances that are unusual and do not conform to established patterns. It is essentially the opposite of clustering, which groups similar data points together.1

A common application of anomaly detection is in fraud detection. While fraud might seem like a problem for predictive modeling, the rarity of fraudulent transactions and the rapid evolution of criminal tactics mean that predictive models can quickly become inaccurate. Anomaly detection focuses on understanding normal behavior to flag unusual transactions. This technique is also used in monitoring systems, such as for intrusion detection. An anomaly, in general terms, is something that deviates from what is ordinary or expected.12