The link-prediction problem for social networks
Journal of the American Society for Information Science and Technology
Fast direction-aware proximity for graph mining
Proceedings of the 13th ACM SIGKDD international conference on Knowledge discovery and data mining
Measuring and extracting proximity graphs in networks
ACM Transactions on Knowledge Discovery from Data (TKDD)
Find me if you can: improving geographical prediction with social and spatial proximity
Proceedings of the 19th international conference on World wide web
Using friendship ties and family circles for link prediction
SNAKDD'08 Proceedings of the Second international conference on Advances in social network mining and analysis
Supervised random walks: predicting and recommending links in social networks
Proceedings of the fourth ACM international conference on Web search and data mining
Inferring social ties across heterogenous networks
Proceedings of the fifth ACM international conference on Web search and data mining
Theoretical justification of popular link prediction heuristics
IJCAI'11 Proceedings of the Twenty-Second international joint conference on Artificial Intelligence - Volume Volume Three
How Visibility and Divided Attention Constrain Social Contagion
SOCIALCOM-PASSAT '12 Proceedings of the 2012 ASE/IEEE International Conference on Social Computing and 2012 ASE/IEEE International Conference on Privacy, Security, Risk and Trust
| Hi-index | 0.00 |
Link prediction is used by many applications to recommend new products or social connections to people. Link prediction leverages information in network structure to identify missing links or predict which new one will form in the future. Recent research has provided a theoretical justification for the success of some popular link prediction heuristics, such as the number of common neighbors and the Adamic-Adar score, by showing that they estimate the distance between nodes in some latent feature space. In this paper we examine the link prediction task from the novel perspective of network flows. We show that how easily two nodes can interact with or influence each other depends not only on their position in the network, but also on the nature of the flow that mediates interactions between them. We show that different types of flows lead to different notions of network proximity, some of which are mathematically equivalent to existing link prediction heuristics. We measure the performance of different heuristics on the missing link prediction task in a variety of real-world social, technological and biological networks. We show that heuristics based on a random walk-type processes outperform the popular Adamic-Adar and the number of common neighbors heuristics in many networks.