Microscale Prediction of Near-Future Crime Concentrations with Street-Level Geosurveillance

摘要

This article proposes a new type of geosurveillance method for monitoring elevated crime activities recorded at the disaggregate street address level. It is a prospective method that combines on a recently developed retrospective method for network-based space-time hot spot detection with a concept used in syndromic surveillance in epidemiology. This method detects emerging concentrations of crime activities at the street level by repeatedly sweeping across a street network using a flexible search window as new incidents are reported. Empirical analysis of drug incident data using a set of search windows with the same spatial extent but different temporal durations suggests that, while all window sizes raise an alarm against a sudden outburst of crime activities, the window with a longer temporal duration is more effective in the early detection of hot spots that are recurrent in nature as well as those that are slow informing a concentration. A distribution of simulated hot spots is also used for examining the performance of the method in the form of days to detect. It shows that searches with a shorter temporal window can furnish a better performance in detecting hot spots that exhibit a sudden outburst with no recurrent pattern.