Capturing On-Line Social Network Link Dynamics Using Event-Driven Sampling

This paper examines the application of an event-driven sampling (EDS) approach to the LiveJournal social network. The intention of the EDS approach is to capture both the content of user blogs in conjunction with their friendship link dynamics over time with a time step of a single day and with high accuracy. The EDS approach makes use of the "always on" Atom feed provided by LiveJournal that contains all public blog posts in near real-time to inform the sampling process of user friendship networks. This has the effect of targeting sampling towards the public active users of the network. We show that the EDS approach is capable of maintaining 98% daily accuracy across all user friendship link dynamics for the class of users that are both public and active. We show that the group of public active users represents approximately 85% of the active network mass. Analysis shows that the network model maintains both small-world and scale-free properties. Data used for the analysis of the EDS technique spans a period of seven months and involves the analysis of data from 4.8 million users and approximately 34 million friendship links. To our knowledge our study is the first to look at and analyze the use of an ldquoalways onrdquo Atom feed like the one provided by LiveJournal to inform a sampling process targeted at capturing user blogs in conjunction with user link dynamics over time within the context of an on-line social network.