|
Activity Number:
|
319
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #304191 |
|
Title:
|
Anomaly Detection Using Scan Statistics on Time Series Hypergraphs
|
|
Author(s):
|
Youngser Park*+ and Carey E. Priebe and David Marchette and Abdou Youssef
|
|
Companies:
|
Johns Hopkins University and Johns Hopkins University and Naval Surface Warfare Center and The George Washington University
|
|
Address:
|
3400 N. Charles St., Baltimore, MD, 21218,
|
|
Keywords:
|
anomaly detection ; link analysis ; social network ; statistical methods ; temporal data
|
|
Abstract:
|
We present a theory of scan statistics on hypergraphs and apply the methodology to a time series of email data. This approach is of interest because a hypergraph is better suited to email data than a graph. This is due to the fact that a hypergraph can contain all the recipients of a message in a single hyperedge rather than treating each recipient separately in a graph. The result shows that scan statistics on hypergraphs can detect certain anomalies that are not apparent by using scan statistics on graphs. We will discuss our methodology in detail and provide an example of anomaly detection using this technique on a time series of Enron email data.
|
