Using kernel methods to visualise crime data

Abstract

Kernel smoothing methods can be used to visualise complex point event data that are distributed geographically. The visualisations generated by these methods allow users to observe trends in the data which may not be apparent from numerical summaries or by visualising the point data themselves. Crime data in the UK has recently been made available to the public for the smallest postal delivery areas in the country (postcodes). Visualisations of the number of recorded events at each postcode for a given area are provided online. In this paper we apply some widely used kernel smoothing methods to this data, to assess the use of kernel smoothing methods as applied to open access data. Kernel smoothing methods are applied to crime data from the Greater London metropolitan area using freely available packages for the open source R program. We also investigate the utility of using simple methods to smooth the data over time. The kernel smoothers used are demonstrated to provide effective and useful visualisations of the data, and it is proposed that these methods should be more widely applied in exploratory data visualisation in the future.