Inferring Social Relationships from Mobility Patterns of Location Based Online Social Networks' Users

Abstract

It is commonly accepted, that we mostly tend to make social ties with people who live near us and whom we meet frequently. However, it is not the case for online social networks' users, who form strong friendship ties with individuals living thousands kilometers away. Obtaining information about users' online experience makes it possible to study the relationship between geographical location and friendship probability. Online social networks give a wide opportunity to collect information about user's geographical location and social ties among users. Check-ins' records can be deployed to extract user's mobility patterns, for study the regularity of user's movements and for the analysis of existence of friendship ties among participants of location based social network. The thesis aims to investigate to what extent the mobility-based information can be useful to make a conclusion about existence of social ties between two given users, by extracting the features, based on their spatial and temporal proximity as well as introducing additional set of features measuring the diversity of visits the given location by unique users. Basing on users mobility features a model for predicting social relationships between users is proposed. One of the possible applications of this model is an analysis of individuals' online and offline social behavior and it can also be deployed in friend recommendation systems.

Date of Award	2013
Original language	American English