Crime in Philadelphia: Bayesian Clustering with Particle Optimization

Accurate estimation of the change in crime over time is a critical first step towards better understanding public safety large urban environments. Bayesian hierarchical modeling natural way to study spatial variation dynamics at neighborhood level, since it facilitates principled sharing information between spatially adjacent neighborhoods. Typically, however, cities contain many physical and social boundaries that may manifest as discontinuities patterns. In this situation, standard prior choices often yield overly-smooth parameter estimates, which can ultimately produce miscalibrated forecasts. To prevent potential over-smoothing, we introduce partitions set neighborhoods into several clusters encourages smoothness within each cluster. terms model implementation, conventional stochastic search techniques are computationally prohibitive, they must traverse combinatorially vast space partitions. We an ensemble optimization procedure simultaneously identifies high probability by solving one problem using new local strategy. then use identified estimate trends Philadelphia 2006 2017. On simulated real data, our proposed method demonstrates good partition selection performance. Supplementary materials for article available online.