Using traffic flow data to analyze the large-scale social networking behavior

The rapid developments of Internet are reshaping many of our routine daily activities, ranging from how we communicate with our friends to how we shop. Understanding the human behaviors and dynamics in such a large-scale social network is essential for better system design, service provisioning, and network management. In this paper, based on real traffic flow data collected from operational networks of Internet Service Provider, the flow graph is constructed to model the interactions among users. We focus on three of the most popular interactive applications on Internet, QQ, Skype and MSN, and investigate the statistical characteristics of the flow graphs. The degree distribution and strength distribution of the flow graphs are analyzed, and the community structure is also discussed. Comparative analysis results show different characteristics of the three interactive applications, which imply the heterogeneous user behaviors on Internet.