Enhanced scan statistic with tightened window for detecting irregularly shaped hotspots

In spatial point data, a hotspot is defined as a group of points with a significantly higher density within an arbitrarily shaped area. Among existing hotspot identification methods, spatial scan statistic, known for its simple mechanism in locating hotspots, has been extensively studied and applied in various fields. However, existing methods rely on pre-defined scanning window shapes, e.g. generic geometries like circles or pre-divided regions, like administrative divisions, and thereby may not accurately capture the irregular shapes of hotspots. This study enhances the spatial scan statistic by introducing a tightened window, which is defined as the window tightened to align with the hotspot’s shape. In our method, without the necessity of outlining the exact geometry, the area of the tightened window, estimated using the nearest distance statistics, is used for calculating the objective function. Experiments with simulated data demonstrate that our method outperforms existing methods in terms of testing hotspots’ significance, identifying arbitrarily shaped hotspots, estimating hotspots’ spatial extent, and reducing subjectivity in parameter selection. An empirical study using taxi pick-up point data shows our method can identify regions with high taxi demand and potential traffic congestion, including subway exits and commercial streets.