基于时空相似度聚类的热点载客路径挖掘

出租车的载客轨迹直接体现了车辆的行驶状态和居民的出行规律，热点载客路径的挖掘为交通管理与规划，居民行为模式发现及出租车载客推荐等具有重要价值. 本文以兰州市3 000 辆出租车载客轨迹为研究对象，提出了基于时空相似性聚类的热点载客路径挖掘算法. 首先，根据出租车的GPS轨迹数据提取出载客轨迹及其核心轨迹；然后，根据提出的相似性度量算法计算核心轨迹的空间相似性、时间相似性及时空相似性，并结合DBSCAN聚类算法对载客轨迹进行聚类；最后，根据聚类结果获取城市热点载客路径的空间分布，并分析了其在工作日和非工作日的差异. 实验结果表明，本文提出的挖掘算法能有效、快速地发现城市热点载客路径的分布.

Hot Passenger Routes Mining Based on Spatial-temporal Similarity Clustering

The taxi passenger trajectory can be exploited to discover the vehicle running state and the law of the travel behaviors of urban citizens. The mining of hot passenger routes has important value for traffic management and planning, citizens’behavior pattern discovery and taxi passenger recommendation. In this paper, a mining algorithm of hot passenger routes based on spatial-temporal similarity clustering is proposed from taxi passenger trajectory generated by over 3 000 taxis for one week in Lanzhou, China. Firstly, the passenger trajectory and its core trajectory are extracted according to the GPS trajectory data of taxi. Then, the spatial similarity, temporal similarity and spatial- temporal similarity of the core trajectory are calculated based on the proposed similarity measurement algorithm. The passenger trajectory is clustered using the DBSCAN clustering algorithm. Finally, the spatial distribution of hot passenger routes is obtained according to the clustering results. The differences of hot passenger routes between weekday and weekend are analyzed. Experimental results show that the proposed mining algorithm can effectively and quickly find the distribution of hot passenger routes.