Applications of Bayesian nonparametric mixture models

Data classification, or clustering, is a complex problem, often addressed using mixture models. In this presentation, I will focus on nonparametric Bayesian mixture models. These models offer several advantages, including flexibility, efficiency in processing small data sets, and the ability to provide uncertainty quantification and simultaneous estimation of density and clustering. I will present two applications of these models to two concrete problems. The first is an ecotoxicological problem, the objective of which is to assess the risks of water contamination by chemicals. Finally, I will illustrate the possibilities offered by these models in the field of criminal networks. The goal is to determine the structure of groups within a mafia organization by studying jointly two networks of co-attended meetings.