Data-Based Detection of Potential Terrorist Attacks: Statistical and Graphical Methods

摘要

This report describes results of research contained in published papers and technical reports on statistical methods in the identification of rare events such as those that arise in terrorist attacks. Because data for national security operations are highly classified, techniques are developed on massive data containing rare events that arise in particle physics, internet communications, and genomics. Results include: (a) useful graphical displays; (b) conditioning high-dimensional data; (c) penalized regression estimators to filter out irrelevant variables.