Coupled Mobility of Social Ties
Figure: For each mobile phone user, we keep track of (A) how many visits are made to locations across the city and (B) construct a social network by tracking calls to others. We can then define (C) the geographic cosine similarity between two users by computing the cosine of the angle between any two vectors in the location space.
We have discovered ubiquitous patterns of similarity in the mobility among social contacts. We show that these features can help us to classify social relationships in data passively collected. Introducing measures of mobility similarity and predictability, we find that that a user’s social ties contain information required to reconstruct his or her visitation patterns very effectively. The hourly and weekly variation in the mobility similarity allows us to classify social ties into three main categories related to social different roles such as co-workers or acquaintances. We present a simple extension to existing mobility models to account for this key characteristics of the mobility similarity observed in social networks. These results add much needed context to social networks generated from big anonymous data sets.
Source: Coupled Human Mobility and Social Ties , pdf, 2014.