Anomalies also referred to as outliers, novelties, noise, deviations, and exceptions are typically defined as the identification of rare items, events, or observations within a pool or set of data that raise suspicions by differing significantly from the majority of the data.

Why should so much importance be placed on anomalies and their detection?

Because anomalies in data will almost always translate to some kind of problem, such as fraud, a defect, medical problems, or errors in a text.

Anomaly detection is a technique used to recognize unusual patterns that do not conform to expected behavior, called outliers. In order to locate anomalies, you need to understand that can fall into several broad categories.

Typically, we consider anomalies to be either point, contextual, or collective in nature. Point anomalies are what you may guess: a single point of data that is very different from the rest. Contextual anomalies are when seemingly good data is only good within a certain context. Collective anomalies are where you consider data in a collective set an anomaly.