Abstract: With broad applications of all kinds of smart phone APP containing Location Based Services（LBS）, a large amount of trajectory data left behind tourists can be collected, and mining hot spots from such data will contribute to intelligent service and emergency management. A clustering-based approach for discovering hot spots in spatial trajectories is proposed. Firstly, the DBSCAN（Density-Based Spatial Clustering of Applications with Noise） algorithm is applied to spatial clustering due to its simplicity, robustness against noise and ability to discover clusters of arbitrary shapes. However, DBSCAN is sensitive to its two initial input parameters and it is hard to determine them a priori. This paper presents an improved method to choose DBSCAN parameters automatically according to the data statistics distribution. Experimental results are obtained from two-dimensional artificial data sets, four-dimensional UCI Iris data sets and stay points data sets. The final results show that the improved algorithm gets good results with respect to the original DBSCAN and k-means algorithms. Finally, the Getis-Ord Gi* hot spot analysis and mapping based on clustering results are conducted in ArcGIS software, and the grading heats of different scenic spots are given, which are equivalent to the information held by the tourism management.

Key words: stay point, spatial clustering, hot spot analysis, Density-Based Spatial Clustering of Applications with Noise（DBSCAN） algorithm, trajectory, scenic spot

摘要： 各种集成位置服务（LBS）的社交和旅游类APP的广泛应用，产生了大量轨迹空间数据，利用这些轨迹数据挖掘游客聚集密度高的热门景点区域，对景区的智慧服务和应急管理具有重要意义。为此，提出了一种基于轨迹停留点空间聚类的景区热点分析方法。重点研究了聚类速度快、能处理噪声、可以发现空间任意形状聚簇的DBSCAN算法，针对其参数需人工选择的不足，提出了一种根据数据统计分布特性来自适应确定参数的改进方法。分别采用人工合成二维数据集、四维Iris真实数据集和景区轨迹停留点三种不同的数据进行了DBSCAN聚类分析及对比实验，结果表明该方法可以自动产生合理的聚簇划分，优于传统DBSCAN和k-means等算法。最后，依据轨迹停留点的空间聚类结果，在ArcGIS软件中实现Getis-Ord Gi*热点分析与制图，并依据分析结果对不同旅游景点进行热度分级，形成的热门景点分布与景区掌握的实际热度信息基本一致，证实了提出方法的有效性。

关键词: 停留点, 空间聚类, 热点分析, DBSCAN算法, 轨迹, 景区