Files
Abstract
We use an undirected graphical model (Markov random field) as a spatial framework to estimate mismanaged waste distributions in a district of Hồ Chí Minh City, Vietnam using data collected throughout the district. The Markov random field allows us to describe joint probabilities of different locations or nodes within the graph to describe possible mismanaged waste distributions. A Bayesian perspective is used within the model where expert information of varying socioeconomic conditions is used as the basis for the nodes' prior distributions and empirical data is used to update the nodes via an auto-model. Once spatial dependencies are accounted for and the nodes' distributions are updated, the final graph is used to simulate possible waste counts and their proportions for the entire district.