Bayesian nonparametric models for spatial data analysis

Abstract

This dissertation proposes and explores several probabilistic models that deal with spatial data in various fields. In particular, Bayesian nonparametric priors are used to solve the problem of parameter selection occurred by density estimation or clustering problem. In the first chapter, we discuss a density estimation model for mobile location data with different volume and measurement errors. In the second chapter, we propose a Bayesian nonparametric joint mixture model for clustering spatially correlated time series based on both spatial and temporal similarities. Finally, in the third chapter, we propose a deep generative cluster model in order to cluster similar patterns of wafers in semiconductor processes. We show potential usefulness of the proposed probabilistic models for a lot of real spatial data through various examples and experimental results that appropriately reflect the nature of the spatial data and the problem situations.