Title: Biclustering bipartite networks via a finite mixture approach Authors:  Dalila Failli - University of Florence (Italy) [presenting]
Maria Francesca Marino - University of Florence (Italy)
Francesca Martella - La Sapienza University of Rome (Italy)
Abstract: Bipartite networks are particularly useful for representing relationships between disjoint sets of nodes, called sending and receiving nodes. Mixtures of latent trait analyzers are modified to perform joint clustering of sending and receiving nodes and to model the dependence between receiving nodes via a latent trait. Thus, the proposed model can partition the data matrix into homogeneous blocks, as in the biclustering approach, and capture the latent variability of network connections within each block, as in the latent trait framework. The proposal also admits the inclusion of nodal attributes in the latent layer of the model to understand how these influence cluster formation. An EM algorithm estimates the model parameters, and Gauss Hermite quadrature is adopted to approximate the likelihood function. The model's parameter recovery and clustering performance are evaluated through a simulation study with a different number of nodes and partitions. In addition to the simulation study, the proposed method is also shown on a real data application.