Link prediction for partially observed networks
This addresses a fundamental problem in network analysis, particularly in fields like genetics, but it is incremental as it builds on existing supervised learning approaches.
The paper tackles link prediction in partially observed networks where no negative examples of absent edges exist, by developing a method that treats the observed network as a sample with different rates for positive and negative examples, and it performs well empirically, including in sparse settings.
Link prediction is one of the fundamental problems in network analysis. In many applications, notably in genetics, a partially observed network may not contain any negative examples of absent edges, which creates a difficulty for many existing supervised learning approaches. We develop a new method which treats the observed network as a sample of the true network with different sampling rates for positive and negative examples. We obtain a relative ranking of potential links by their probabilities, utilizing information on node covariates as well as on network topology. Empirically, the method performs well under many settings, including when the observed network is sparse. We apply the method to a protein-protein interaction network and a school friendship network.