An extension of finite mixtures of latent trait analyzers for biclustering bipartite networks

Network data analysis has received increasing attention recently. Bipartite networks represent a specific type of network data describing the relationships between disjoint sets of nodes, called sending and receiving nodes. We extend the Mixture of Latent Trait An alyzers (MLTA) specifically tailored for the analysis of bipartite networks to achieve a twofold goal. First, the aim is to perform a joint clustering of sending and receiving nodes, thus partitioning the data matrix into homogeneous blocks, as in the biclustering approach. In addition, a latent trait is used to model the dependence between receiving nodes, as in the latent trait framework. The proposal also admits the inclusion of nodal attributes on the latent layer of the model to understand how they affect cluster formation. An EM algorithm with Gauss Hermite approximation is proposed to estimate the model parameters.