Traditional methods of crime mapping, such as kernel density estimation (KDE), are used to generate risk surfaces that indicate where crime has previously clustered. As such, they consider the location of crime events but ignore their timing. Inspired by the above findings, Bowers et al. (2004) proposed a method of predictive crime mapping, named ProMap, that models the way in which crime clusters (or appears to spread) in space and time. To do this, the expected risk at a location for a particular period (usually the next day, few days or the next week) is estimated as a function of the density of crime that has occurred at or near to that location. However, events are also inversely weighted according to when they occurred, so that more recent crimes receive a greater weighting. The simplest form of the function sums the product of inverse time and distance weights given to each crime in the data set for the locations of interest (e.g. a series of grid cells).