Mining anomalies in graph data

Abstract

Mining graph data is an important data mining task due to its significance in network analysis and several other contemporary applications. With this backdrop, this chapter explores the potential applications of outlier detection principles in graph/network data mining for anomaly detection. One of the focus areas is to detect arbitrary subgraphs of the input graph exhibiting deviating characteristics. In this direction, graph mining methods developed based on latest algorithmic techniques for detecting various kinds of anomalous subgraphs are explored here. It also includes an experimental study involving benchmark graph data sets to demonstrate the process of anomaly detection in network/graph data.