Performance Evaluation of Temporal Anomaly Detection on Simulated Social Networks Using a Moving Window-Based Scan Method

• Activity Number: 138
• Type: Contributed
• Date/Time: Monday, August 1, 2016 : 8:30 AM to 10:20 AM
• Sponsor: Social Statistics Section
• Author(s): Meng Zhao* and William H. Woodall and Anne R. Driscoll and Ronald D. Fricker and Dan Spitzner
• Companies: Virginia Polytechnic Institute and State University and Virginia Polytechnic Institute and State University and Virginia Polytechnic Institute and State University and Virginia Polytechnic Institute and State University and University of Virginia
• Keywords: Networks ; Performance ; Scan Method ; Binomial ; Moving Window ; ATOS

Abstract

Timely detection of anomalous events in networks, particularly social networks, is a problem of increasing interest and relevance. A variety of methods have been proposed for monitoring such networks, including the window based scan method for social network monitoring proposed by Priebe et al. (2005). However, research assessing the performance of this and other methods has been sparse. In this article, we utilize simulated social network structures to study the performance of Priebe et al.'s method. We show that it frequently cannot detect anomalous events in a network even under very simple conditions. The detection power tends to be adequate only when more than half of the social network experiences anomalous behavior. Simulation studies are employed to show that improved detection rate and shortened monitoring delays can be achieved by lagging the moving window used for standardization, lowering the signaling threshold and using shorter moving windows at the initial stage of monitoring.