In this work, we consider time series of daily concentrations of PM10 monitored in Lombardia and Emilia-Romagna during 2018. With the aim of clustering those spatial time series, we propose a Bayesian nonparametric mixture of autoregressive processes and assume as mixing measure a spatial product partition model. We focus on the implementation of this model into BayesMix, a new C++ library for Bayesian inference on nonparametric mixture models via Markov Chain Monte Carlo. The main feature of this library is its extensibility, which guarantees a seamless integration of new classes of mixture models, like the one we introduce in this paper, without compromising efficiency.
Model-Based Clustering of Spatial Time Series Through the BayesMix library
Gianella, Matteo;Guglielmi, Alessandra
2025-01-01
Abstract
In this work, we consider time series of daily concentrations of PM10 monitored in Lombardia and Emilia-Romagna during 2018. With the aim of clustering those spatial time series, we propose a Bayesian nonparametric mixture of autoregressive processes and assume as mixing measure a spatial product partition model. We focus on the implementation of this model into BayesMix, a new C++ library for Bayesian inference on nonparametric mixture models via Markov Chain Monte Carlo. The main feature of this library is its extensibility, which guarantees a seamless integration of new classes of mixture models, like the one we introduce in this paper, without compromising efficiency.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
