Inferring networks of diffusion and influence

by Gomez Rodriguez, M.; Leskovec, J. & Krause, A.

In order to study network diffusion, we need to

1. identify the contagion (idea, information, virus, phrase),
2. track the contagion,
3. infer the diffusion network based on the observed node/time of infection-pattern.

This article discusses an algorithm for inferring diffusion networks based on this information and presents experiments based on

1. simulated network structures, and
2. a dataset containing more than 172 million news articles from 1 million online sources mirrored between September 2008 and August 2009.

The authors use MemeTracker to identify more than 343 million textual phrases (=contagion) to track over time.