Link Prediction in Human Contact Networks Using Online Social Ties

Human contact traces has allowed researchers to study how mobile users contact each other in different environments. Most of these traces include other types of useful information such as users´ social profiles and their online social connections. This explicit social information is important since it can be exploited for augmenting the knowledge of user behavior and hence improve the quality of human mobility analysis. In this paper, we use online social connections for predicting user´s contacts. Specifically, we study the prediction of links in human contact networks as a graph inference problem, where the existence of an edge is predicted using different similarity measures between nodes. First, we predict the edges of the contact graph when we have only information about users´ online social network. Next, we analyze the effectiveness of using both the online social network and a part of the contact network for contact prediction. In both settings, our results show that Resource Allocation plays a significant role in contact prediction. Finally, we show the importance of online social similarity in identifying stronger ties.