Centrality Measures of Dynamic Social Networks

Abstract

Modern criminal networks are constantly changing to maintain secrecy, recruit members, and coordinate activities. Attempts to uncover important elements of these networks need to incorporate dynamic trends to provide useful findings and disrupt harmful plans. Our research provides a promising approach whereby analysts can forecast network behavior and stay a step ahead of their adversaries. This report explores the theoretical background of dynamic networks and uses the network measures of degree, closeness, betweenness, and eigenvector centrality over time to conduct network trend analysis. As a case study, I examined the Ali Baba data set that provides messages from a fictitious terrorist cell over a seven-month period. The force-directed Fruchterman-Reingold algorithm was used to visualize the Ali Baba network each month to identify structure, distinguish key players, and understand behavioral roles. Despite the low density of interactions, results revealed the ranking of eigenvector centrality to match the terrorist attack cycle. Several methods for centrality measure prediction are also evaluated, including regression and moving average. Lastly, the results of the removal of a key node from a scale-free criminal network are examined. These examples are an important step in the continuing effort to predict terrorist network behavior.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Nov 01, 2012
Accession Number
ADA571973

Entities

People

  • Allison Moore

Organizations

  • United States Army Research Laboratory

Tags

Communities of Interest

  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Algorithms
  • Case Studies
  • Computer Science
  • Criminals
  • Data Mining
  • Data Sets
  • Economic Forecasting
  • Eigenvectors
  • Equations
  • Information Science
  • Mathematics
  • Network Science
  • Personnel Management
  • Polynomials
  • Social Networks
  • Statistics
  • Terrorists

Fields of Study

  • Computer science

Readers

  • Economics
  • Neural Network Machine Learning.