Anomaly detection in banking operations

This paper presents an overview of anomaly detection algorithms and methodology, focusing on the context of banking operations applications. The main principles of anomaly detection are first presented, followed by listing some of the areas in banking that can benefit from anomaly detection. We then discuss traditional nearest-neighbor and clustering-based approaches. Time series and other sequential data analysis approaches are described. The problems posed by categorical data are also discussed, along with the methods proposed in the literature to address the same. The ensemble methods are presented, followed by mathematical perspectives on anomaly detection.