Quantitative Methods in Defense and National Security 2007

Predicting Unobserved Links in Covert Networks
David Marchette, (Naval Surface Warfare Center), dmarchette@gmail.com


In covert networks it is often the case that some links will be unobserved: it will be unknown whether potential edges are present in the graph or not. I will discuss a simple generative model for social networks, the constrained random dot product graph (CRDPG), and illustrate how this can be used to suggest high probability edges for further investigation. The model will be illustrated on a dataset consisting of alliances between nations. These data have covariates on the vertices which will be used to improve the performance of the prediction. Other datasets will be discussed as time permits.

Take me back to the main conference page.