Link prediction and affiliation recommendation using auxiliary sources of information

Berkant Savas
Department of Science and Technology
Campus Norrköping
Linköping University


Link prediction and affiliation recommendation are fundamental problems in social network analysis and modern-day commercial applications. Most existing research approaches this problem by exploring the topological structure of a social network using only one source of information. However, in many application domains, in addition to the social network of interest, there are a number of auxiliary networks and/or derived proximity networks available. In this talk we will present a few ideas on how to utilise auxiliary sources of information for the task at hand. In particular we will present: (1) a supervised learning framework that can learn the dynamics of social networks in the presence of auxiliary networks; (2) a feature design scheme for constructing a rich variety of path-based features using auxiliary sources, and (3) an effective feature selection strategy based on structured sparsity. Experiments on real-world networks show that our models can effectively learn to predict new links/recommendation using auxiliary sources, yielding higher prediction accuracy than unsupervised and single source supervised models.