A data-driven adaptive geospatial hotspot detection approach in smart cities

Hotspot detection from geo-referenced urban data is critical for smart city research, such as traffic management and policy making. However, the classical clustering or classification approach for hotspot detection mainly aims at identifying "hotspot areas" rather than specific points, and the setting of global parameters such as search bandwidth can lead to inaccurate results when processing multi-density urban data. In this article, a data-driven adaptive hotspot detection (AHD) approach based on kernel density analysis is proposed and applied to various spatial objects. The adaptive search bandwidth is automatically calculated depending on the local density. Window detection is used to extract the specific hotspots in AHD, thus realizing a small-scale characterization of urban hotspots. Through the trajectory data of Harbin City taxis and New York City crime data, Geo-information Tupu is used to analyze the obtained specific hotspots and verify the effectiveness of AHD, providing new ideas for further research.