Outlier Detection

Synonyms

Anomaly detection

Definition

Although there is no consensus on how to describe outliers, Barnet’s definition is accepted by many statisticians and computer scientists, describing an outlier as “one observation that appears to deviate markedly from other members of the sample in which it occurs (Barnett and Lewis 1994).” In recent years, outlier detection has begun to be widely used in numerous applications. In different applications, outliers may have different names such as anomalies, deviations, exceptions, faults, and irregularities. Outlier detection can help identify intrusions in computer networks, locate malfunctioning parts in a manufacture streamline, pinpoint suspicious usage of credit cards, and monitor unusual changes of stock prices.

Main Text

Rare events often reveal more important information than common ones, especially in case of the computer security, emergency responses, and medical diagnoses. For example, a web server is expected to handle thousands of...