Link prediction using probabilistic group models of network structure

Authors:
Akira Ninagawa;Koji Eguchi
Affiliations:
Kobe University, Nada-ku, Kobe, Japan;Kobe University, Nada-ku, Kobe, Japan
Venue:
Proceedings of the 2010 ACM Symposium on Applied Computing
Year:
2010

Citing 4
Cited 0

Probabilistic latent semantic indexing

Proceedings of the 22nd annual international ACM SIGIR conference on Research and development in information retrieval
Latent dirichlet allocation

The Journal of Machine Learning Research
The link prediction problem for social networks

CIKM '03 Proceedings of the twelfth international conference on Information and knowledge management
Supervised enzyme network inference from the integration of genomic data and chemical information

Bioinformatics

Quantified Score

Hi-index	0.00

Visualization

Abstract

Modeling of complex networks is a crucial task such as in biology and social sciences. A large number of researches have been conducted for such a problem; however, most of them require explicit, specific prior knowledge on target networks. On the other hand, a few recent works on multinomial mixture models presented that those models do not require such explicit prior knowledge and turned out to be effective for the task of group detection of vertices such as in social networks. This paper focuses on another task, link prediction in such complex networks, using a Bayesian multinomial mixture model, which assumes unobservable prior distributions over multinomial mixtures based on network structure and are estimated using Bayesian inference via Gibbs sampling. We demonstrate that link prediction performance was significantly improved using this method, compared to five conventional methods, through experiments using a metabolic network and a co-authorship network.