Machine Learning: Anomaly Detection
Myeongsu Kang · 2018
It is important to identify deviation from the nominally healthy behavior of the product and detect the onset of the product's potential faults for achieving prognostics and health management (PHM). This chapter offers a comprehensive overview of the research on anomaly detection and discusses the challenges in anomaly detection. For anomaly detection, methods can be categorized into distance-based, clustering based, classification-based, and statistical anomaly detection methods. The chapter provides the underlying background of the type of anomalies that can be classified into one of the following categories: point anomalies, contextual anomalies, and collective anomalies. Clustering is the partitioning of a dataset into clusters by maximizing inter-cluster distances and minimizing intra-cluster distances. The chapter summarizes the advantages and disadvantages of clustering-based anomaly detection methods. A self-organizing maps (SOM), also known as a Kohonen neural network, is a type of unsupervised learning.