Modeling Blog Dynamics

Abstract

How do blogs produce posts? What local, underlying mechanisms lead to the bursty temporal behaviors observed in blog networks? Earlier work analyzed network patterns of blogs and found that blog behavior is bursty and often follows power laws in both topological and temporal characteristics. However, no intuitive and realistic model has yet been introduced, that can lead to such patterns. This is exactly the focus of this work. We propose a generative model that uses simple and intuitive principles for each individual blog, and yet it is able to produce the temporal characteristics of the blogosphere together with global topological network patterns, like power-laws for degree distributions, for inter-posting times, and several more. Our model ZC uses a novel 'zero-crossing' approach based on a random walk, combined with other powerful ideas like exploration and exploitation. This makes it the first model to simultaneously model the topology and temporal dynamics of the blogosphere. We validate our model with experiments on a large collection of 45,000 blogs and 2.2 million posts.