Top-k link recommendation in social networks
- Publication Type:
- Conference Proceeding
- Proceedings - IEEE International Conference on Data Mining, ICDM, 2016, 2016-January pp. 389 - 398
- Issue Date:
Copyright Clearance Process
- Recently Added
- In Progress
- Closed Access
This item is closed access and not available.
© 2015 IEEE. Inferring potential links is a fundamental problem in social networks. In the link recommendation problem, the aim is to suggest a list of potential people to each user, ordered by the preferences of the user. Although various approaches have been developed to solve this problem, the difficulty of producing a ranking list with high precision at the top - the most important consideration for real world applications - remains largely an open problem. In this work, we propose two top-k link recommendation algorithms which focus on optimizing the top ranked links. For this purpose, we define a cost-sensitive ranking loss which penalizes the mistakes at the top of a ranked list more than the mistakes at the bottom. In particular, we propose a log loss, derive its surrogate, and formulate a top-k link recommendation model by optimizing this surrogate loss function based upon latent features. Moreover, we extend this top-k link recommendation model by incorporating both the latent features and explicit features of the network. Finally, an efficient learning scheme to learn the model parameters is provided. We conduct empirical studies based upon four real world datasets, i.e., Wikipedia, CondMat, Epinions, and MovieLens 1M, of which the largest network contains more than 70 thousand nodes and over one million links. Our experiments demonstrate that the proposed algorithms outperform several state-of-the-art methods.
Please use this identifier to cite or link to this item: