基于聚类算法的城市交通热点挖掘及居民出行时空需求分析——以成都市出租车轨迹数据为例

Mining Urban Traffic Hotspots and Analyzing Spatiotemporal Travel Demand of Residents Based on Clustering Algorithms—A Case Study of Taxi Trajectory Data in Chengdu

摘要: 为精准感知城市交通运行状态并深度洞察居民出行规律，本文以成都市出租车GPS轨迹数据为例，提出了一套基于聚类算法的城市交通“热点路段”挖掘、量化及应用分析框架。研究首先对原始轨迹数据进行预处理，提取高质量的OD (起讫点)数据集；随后，分别采用DBSCAN与K-means算法进行交通热点路段挖掘，并通过轮廓系数、离群点率等多个指标对比验证了DBSCAN算法在识别路网形态热点上的优越性；在此基础上，构建了综合出行规模与空间密度的热度量化模型，对识别出的热点路段进行了定量评估。研究结果表明：1) 成功识别出人民南路三段等多个关键交通热点，并发现其热度与周边功能属性高度相关；2) 居民出行在工作日与休息日呈现显著不同的高峰时段特征，且工作日高峰时段的OD点在空间上与通勤规律高度吻合。本研究不仅为利用交通大数据进行精细化城市分析提供了方法论参考，也为交通资源优化配置、拥堵疏导及城市规划提供了数据驱动的决策支持。

Abstract: To accurately perceive the operational status of urban traffic and gain deep insights into residents’ travel patterns, this paper proposes a framework for mining, quantifying, and analyzing urban traffic “hotspot road segments” based on clustering algorithms, using GPS trajectory data from taxis in Chengdu as an example. The study first preprocesses the raw trajectory data to extract a high-quality OD (origin-destination) dataset. Subsequently, DBSCAN and K-means algorithms are employed to mine traffic hotspot road segments, and the superiority of the DBSCAN algorithm in identifying hotspots in road network patterns is verified through multiple indicators such as the contour coefficient and outlier rate. Based on this, a heat quantification model that integrates travel scale and spatial density is constructed to quantitatively evaluate the identified hotspot road segments. The research results show that: 1) Several key traffic hotspots, such as the third section of Renmin South Road, have been successfully identified, and their heat is highly correlated with surrounding functional attributes; 2) Residents’ travel patterns exhibit significantly different peak period characteristics on weekdays and weekends, and the OD points during peak periods on weekdays are highly spatially consistent with commuting patterns. This study not only provides a methodological reference for utilizing traffic big data for refined urban analysis but also offers data-driven decision support for optimizing traffic resource allocation, easing congestion, and urban planning.