|
Abstract:
|
We present a statistical method to generate a series of predicted networks based on the historical evolution of social relations in a given population. This allows an investigator to incorporate and model uncertainty, which is inherent in prediction. The method, which is based on a novel and flexible procedure to sample dynamic networks given a probability distribution on evolving network properties, permits the use of a broad class of approaches in order to model trends, seasonal variability, uncertainty, and changes in population compositions. The proposed method is demonstrated on two dynamic networks; the first represents the sponsor/co-sponsor relationships between senators indicated from bills introduced in the US Senate from 2003-2012. The second is temporal network data-sampled every 20 seconds-representing interactions between participants of the ACM Hypertext 2009 conference. The proposed method enables investigators to make rapid and informed decisions regarding network interventions such as mechanisms to either encourage information diffusion or minimize the impact of a contagion; we present a network-based intervention for the ACM conference for demonstration.
|
