A Bayesian semiparametric model for terrorist networks

A recent field of research employs network-analysis' tools to the \emph{dark network} framework, in which pairwise informations about terrorists' activities are available. In this work we focus on the ``Noordin Mohamed Top'' dataset, developing an asymmetric approach that treats one network as response and the remaining as covariates. We aim to identify which information may be useful in predicting terrorists' collaboration in a bombing attack, identifying at the same time the most influential subjects involved in these dynamics. Such aim is addressed with a Bayesian semi-parametric model for networks that, through a suitable prior specification, integrates a flexible regularization and the detection of leading nodes. Taking advantage of the P\'{o}lya-Gamma data augmentation scheme, we develop an efficient Gibbs sampler to make inference on the parameters involved