Acquaintance or partner? Predicting partnership in online and location-based social networks

Abstract

Existing approaches to predicting tie strength between users involve either online social networks or location-based social networks. To date, few studies combined these networks to investigate the intensity of social relations between users. In this paper we analyzed tie strength defined as partners and acquaintances in two domains: a location-based social network and an online social network (Second Life). We compared user pairs in terms of their partnership and found significant differences between partners and acquaintances. Following these observations, we evaluated the social proximity of users via supervised and unsupervised learning algorithms and established that homophilic features were most valuable for the prediction of partnership.