Model-based Clustering for Multivariate Networks
share
Network data are relational data recorded among a group of individuals, the nodes. Multiple relations observed among the same set of nodes may be represented by means of different networks, using a so-called multidimensional network, or multiplex. We propose a latent space model for network data that enables clustering of the nodes in a latent space, with clusters in this space corresponding to communities of nodes. The clustering structure is modelled using an infinite mixture distribution framework, which allows to perform joint inference on the number of clusters and the cluster parameters. The method is tested on simulated data experiments and is shown in application to a multivariate network among students.
READ FULL TEXT