Real-time Anomaly Detection at Scale

Abstract

Anomaly Detection is a field of Machine Learning used by systems to identify observations that differ from the majority of data and are often linked directly to unexpected behaviour, errors, and other forms of novelties. Common applications of Anomaly Detection include but are not limited to credit card fraud detection, machinery or computer behaviour monitoring, network intrusion detection, real-time analytics, etc. In this thesis we study algorithms and methods for Anomaly Detection that enable identification of outliers both in real-time, in order to prevent unwanted events as soon as possible, and at a big scale, since the volume of data in our era is growing exponentially.