Abstract

Nowadays many people are members of multiple on- line social networks simultaneously, such as Facebook, Twitter and some other instant messaging circles. But these networks are usually isolated from each other. Mapping common users across these social networks will benefit many applications. Methods based on user- name comparison perform well on parts of users, how- ever they can not work in the following situations: (a) users choose different usernames in different networks; (b) a unique username corresponds to different individ- uals. In this paper, we propose to utilize social struc- tures to improve the mapping performance. Specifi- cally, a novel subspace learning algorithm, Manifold Alignment on Hypergraph (MAH), is proposed. Dif- ferent from traditional semi-supervised manifold align- ment methods, we use hypergraph to model high-order relations here. For a target user in one network, the proposed algorithm ranks all users in the other net- work by their possibilities of being the corresponding user. Moreover, methods based on username compari- son can be incorporated into our algorithm easily to fur- ther boost the mapping accuracy. Experimental results have demonstrated the effectiveness of our proposed al- gorithm in mapping users across networks.