An Evaluation Method for Unsupervised Anomaly Detection Algorithms

Abstract

In data mining, anomaly detection aims at identifying the observations which do not conform to an expected behavior. To date, a large number of techniques for anomaly detection have been proposed and developed. These techniques have been successfully applied to many real world applications such as fraud detection for credit cards and intrusion detection in network security. However, there are very little research relating to the method for evaluating the goodness of unsupervised anomaly detection techniques. In this paper, the authors introduce a method for evaluating the performance of unsupervised anomaly detection techniques. The method is based on the application of internal validation metrics in clustering algorithms to anomaly detection. The experiments were conducted on a number of benchmarking datasets. The results are compared with the result of a recent proposed approach that shows that some proposed metrics are very consistent when being used to evaluate the performance of unsupervised anomaly detection algorithms.