A taxonomy and platform for anomaly detection
Gheorghe Sebestyen, Anca Hângan, Zoltan Czako, György Kovács · 2018
There are hundreds of anomaly detection methods developed for different purposes and using a wide range of theoretical backgrounds, from system theory and signal processing towards artificial intelligence techniques. Therefore, it is very difficult for a specialist in a given domain (e.g. finance, industrial engineering, networking, environment monitoring, etc.) to select an anomaly detection method that fits best for a given application. The goal of this paper is to propose a taxonomy for anomaly detection methods and also to present a platform that allows a developer to find and tune a given anomaly detection method that is optimal for an application. The platform provides the basic functionalities needed to acquire, process and visualize multidimensional data collected from different sources, as well as the means to test and compare different anomaly detection techniques.