Learning Network Dynamics from Tumblr®: A Search for Influential Users

摘要

This work offers an original analysis of a unique data set gathered from the blogging website Tumblr by developing and applying a new data driven method for investigating network dynamics. To our knowledge, this is the first effort to analyze the spread of information on Tumblr on a such a large scale, and our method generally applies to networks where nodes have time-evolving states. We start by testing our method on simulated data, then we follow over 50,000 blogs on Tumblr over a year of activity to determine not only which blogs are influential, but more importantly, how these blogs spread their content.