Spatio-temporal modelling of crime using low discrepancy sequences

Abstract

We perform spatio-temporal modelling of burglary data in order to predict areas of high criminal risk for local authorities. We wish to account for several spatio-temporal factors as latent processes to make the model as realistic as possible, thus creating a model with a large latent eld with several hyperparameters. Analysis of the model is done using Integrated Nested Laplace Approximations (INLA) (Rue et al. 2009), a fast Bayesian inference methodology that provides more computationally efficient estimations than Markov Chain Monte Carlo (MCMC) methods.