2015/08/31 by Desislava Hristova, Anastasios Noulas, Hristova, Desislava +7
Physics and Astronomy · Computer Science · #Complex Network Analysis Techniques #Advanced Graph Neural Networks #Recommender Systems and Techniques
paper · pdf · doi:10.48550/arxiv.1508.07876
Online social systems are multiplex in nature as multiple links may exist\nbetween the same two users across different social networks. In this work, we\nintroduce a framework for studying links and interactions between users beyond\nthe individual social network. Exploring the cross-section of two popular\nonline platforms - Twitter and location-based social network Foursquare - we\nrepresent the two together as a composite multilayer online social network.\nThrough this paradigm we study the interactions of pairs of users\ndifferentiating between those with links on one or both networks. We find that\nusers with multiplex links, who are connected on both networks, interact more\nand have greater neighbourhood overlap on both platforms, in comparison with\npairs who are connected on just one of the social networks. In particular, the\nmost frequented locations of users are considerably closer, and similarity is\nconsiderably greater among multiplex links. We present a number of structural\nand interaction features, such as the multilayer Adamic/Adar coefficient, which\nare based on the extension of the concept of the node neighbourhood beyond the\nsingle network. Our evaluation, which aims to shed light on the implications of\nmultiplexity for the link generation process, shows that multilayer features,\nconstructed from properties across social networks, perform better than their\nsingle network counterparts in predicting links across networks. We propose\nthat combining information from multiple networks in a multilayer configuration\ncan provide new insights into user interactions on online social networks, and\ncan significantly improve link prediction overall with valuable applications to\nsocial bootstrapping and friend recommendations.\n