Bayesian Community Detection with Node Features (306301)Lizhen Lin, University of Notre Dame
*Luyi Shen, University of Notre Dame
Keywords: community detection, node features, distance dependent Chinese Restaurant Process, stochastic block model
Many algorithms have been developed for community detection in networks which is among one of the central tasks in network analysis. However, most of the algorithms do not incorporate node features into clustering which often carries useful additional information. We propose a novel Bayesian model for community detection in networks that can effectively take into account the additional node features for clustering. Our model employs a distance based dependent Chinese restaurant process as a random partition prior which is combined with a block model for inference. Unlike most of the existing algorithms, our model does not require the knowledge of the number of clusters k which is often unknown in reality. Our simulation results show that our model provides a more flexible and accurate estimation of the communities or clustering structure while being able to simultaneously estimate the number of communities.