A0883
Title: Bayesian nonparametric mixture models and clustering for ecological risks
Authors: Louise Alamichel - Universite Grenoble Alpes (France) [presenting]
Julyan Arbel - Inria (France)
Guillaume Kon Kam King - Université Paris-Saclay, INRAE (France)
Igor Pruenster - Bocconi University (Italy)
Abstract: Bayesian nonparametric mixture models are common for modeling complex data. These models are well-known for being consistent when used for density estimation. However, the consistency of the posterior distribution does not provide asymptotic guarantees in the context of clustering problems. Until recently, there has been a lack of asymptotic guarantees regarding the posterior number of clusters for these models. After studying the asymptotic properties of these models, one of them is applied to a real-world problem in ecotoxicology. A Bayesian nonparametric mixture model is proposed to assess the ecological risks of water contaminants. The choice of a Bayesian nonparametric approach offers several advantages, including its efficiency in handling small datasets typical of environmental risk assessments, its ability to provide uncertainty quantification, and its capacity for simultaneous density and clustering estimation. Through systematic simulation studies and analysis of real datasets, the superiority of the Bayesian nonparametric approach is demonstrated over classical methods for this problem.
