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.

中文翻译：

---

智慧城市中数据驱动的自适应地理空间热点检测方法

根据地理参考城市数据进行热点检测对于智慧城市研究（例如交通管理和政策制定）至关重要。然而，用于热点检测的经典聚类或分类方法主要旨在识别“热点区域”而不是特定点，并且搜索带宽等全局参数的设置在处理多密度城市数据时可能导致结果不准确。本文提出了一种基于核密度分析的数据驱动的自适应热点检测（AHD）方法，并将其应用于各种空间对象。自适应搜索带宽根据局部密度自动计算。 AHD中利用窗口检测来提取特定热点，从而实现城市热点的小尺度表征。通过哈尔滨市出租车的轨迹数据和纽约市的犯罪数据，利用地理信息图谱对获取的特定热点进行分析，验证AHD的有效性，为进一步研究提供新思路。