Uncovering the spatiotemporal patterns of traffic congestion from large-scale trajectory data: A complex network approach

Abstract

Understanding the spatiotemporal characteristics of traffic congestion is the cornerstone of generating traffic management and control strategies. Based on the large-scale taxi trajectory data in Shenzhen, China, this study designs an effective framework to explore the spatiotemporal patterns of traffic congestions. To bridge trajectory data with urban road networks, we develop a two-stage map-matching method from the aspects of distance and angle. Then, the free-flow speed of each road segment is extracted and employed to identify traffic congestion. In this way, a novel complex network method, named chronological network (chronnet), is utilized for traffic congestion modeling, and we employ an overlapping community detection algorithm to identify region-level bottlenecks. According to the network properties, we explore the influence scope of traffic congestions and uncover the role of each road segment in the propagation process. Meanwhile, community detection results indicate that there are typical local clustering structures in traffic congestions, and each community also has its unique traffic characteristics. Overall, these findings reveal that the complex network can effectively mine the consecutive patterns of traffic congestion.