Crime Analyses Using R

There has been a considerable rise in the efforts toward predictive policing around the world. We study the backbone of these policing methods by understanding the predictive engines that are behind such efforts. We study, clean, process, and visualize crime data before building a predictive model to estimate the number of crimes expected to happen in the near future in a given area. Given the overdispersed nature of crime counts, we model them using a negative binomial distribution and compare the actual and predicted values on an out-of-time validation set. All processes are completely executed in the R statistical software.