Purpose-Driven Communities in Multiplex Networks: Thresholding User-Engaged Layer Aggregation
Abstract
Discovering true and meaningful communities in dark networks is a non-trivial yet useful task. Because terrorists work hard to hide their relationships/network, analysts have an incomplete picture of their strategy; even worse, the degree of incompleteness is unknown. To better protect our nation, analysts would benefit from a tool that helps them identify meaningful terrorist communities. This thesis introduces a general-purpose algorithm for community detection in multiplex dark networks using the layers of the network based on edge attributes. The methodology includes community detection details from each layer, yet it is still flexible enough to be meaningful in a variety of networks based on the users interest. The aim of this thesis is to build on current layer aggregation methodologies as well as preexisting community detection algorithms. We apply our algorithm to three multiplex terrorist networks: Noordin Top Network, Boko Haram and Fuerzas Armadas Revolucionarias de Colombia (FARC). We validate our algorithm by measuring adjusted conductance and cluster adequacy with respect to community quality. We demonstrate the utility of our community partitions by developing a community guided network shortest path interdiction model, which disrupts the information flow in the Noordin Top Network
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 2016
- Accession Number
- AD1026724
Entities
People
- Ryan E. Miller
Organizations
- Naval Postgraduate School