Hotspot analysis

Hotspot analysis identifies statistically significant spatial clusters, areas where an event or value is unusually high ("hot") or unusually low ("cold"), rather than relying on visual inspection of a map alone. It is widely used in policing to find crime clusters, in public health to find disease clusters, and in urban planning for spatial analysis of service demand.

A common underlying statistic is the Getis-Ord Gi* z-score, which compares each location's value and its neighbors' values to the study area's overall mean, flagging a location as a significant hot spot when its local cluster is more extreme than random spatial distribution would produce. This differs from a simple choropleth map or density surface, which shows where values are concentrated but does not test whether that concentration is statistically meaningful versus chance clustering. The analysis requires geocoded data, run within a geographic information system, often aggregated to a grid or administrative unit.

Hotspot analysis matters because it lets agencies target resources, patrols, inspections, outreach, at the places where an anomaly detection approach would flag genuine clustering rather than noise. Pitfalls include feedback loops, for example "hot spot policing" that increases enforcement can itself raise the recorded crime rate in that area, and sensitivity to the size and shape of the areal units chosen, known as the modifiable areal unit problem, which can change which places appear as hotspots.