Material Detail

This video was recorded at 6th International Workshop on Mining and Learning with Graphs (MLG), Helsinki 2008. We examine the communication patterns of millions of anonymized mobile phone users. Based on call records, we construct a communication network where vertices are subscribers and edge weights are defined as aggregated duration of calls, reflecting the strengths of social ties between callers. We observe a coupling between tie strengths and network topology: at the "local" level, strong ties are associated with densely connected network neighborhoods, providing the first large-scale confirmation of the Granovetter hypothesis. Based on fragmentation analysis, weak ties are seen to play an important role at the network level, accounting for global connectivity. The observed coupling is shown to significantly slow down the spreading of random information, resulting in dynamic trapping of information in communities.

Disciplines:

• Science and Technology  / Computer Science