Wiley Handbook of Science and Technology for Homeland Security, 4 Volume Set

BAYESIAN NETWORKS

DENNIS BUEDE, SUZANNE M. MAHONEY, AND JOSEPH A. TATMAN

Innovative Decisions, Inc., Vienna, Virginia

1 INTRODUCTION

Bayesian networks (BNs) may be used to address Department of Homeland Security (DHS) needs ranging from identifying and tracking suspect individuals to analyzing vulnerabilities and providing warning of attack. BNs factor complex problems into manageable networks of random variables. The graphical interface of BNs facilitates explanation while the local nature of the probabilistic relationships facilitates assessment. Computational algorithms provide rapid evaluation of available evidence to support decision-making in complex and evolving situations.

Section 2 provides a brief introduction to BNs while Section 3 provides some insight into the worldwide research community. The next section describes the application of BNs to meet DHS' critical needs. In the final section, research goals relevant to meet DHS's critical needs are summarized.

2 SCIENTIFIC OVERVIEW

This section presents the basics of BNs. This is followed by a section on inference (or computation). The issue of obtaining the probabilistic inputs for a model is described under knowledge acquisition.

2.1 Modeling with Bayesian Networks

A BNs is a factorization of a joint probability distribution. An acyclic directed graph specifies the network's structure. Nodes stand for random variables. Directed arcs indicate probabilistic dependence. Lack of an arc indicates probabilistic independence. ...