Abstract:With the development of in-vehicle data collection devices, GPS trajectory has become a priority source to identify traffic congestion and understand the operational states of road network in recent years. This study aims to investigate the relationship between traffic congestion and built environment, including traffic-related factors and land use. Fuzzy C-means clustering was used to conduct an exhaustive study on the 24-hour congestion pattern of road segments in an urban area, so that the spatial autoregressive moving average model (SARMA) could be introduced to analyze the output from the clustering analysis to establish the relationship between built environment and the 24-hour congestion pattern. The clustering result classified the road segments into four congestion levels, while the regression explained the impact of 12 traffic-related factors and land-use factors on the road congestion pattern. The continuous congestion was found to mainly occur in the city center, and the factors, such as road type, bus station in the vicinity, ramp nearby, commercial land use, and so on, had large impacts on congestion formation. The Fuzzy C-means clustering is proposed to be combined with quantitative spatial regression, and the overall evaluation process will assist to assess the spatial-temporal levels of service regarding traffic from the congestion perspective.Keywords: Congestion pattern, taxi GPS data, fuzzy C-means clustering, spatiotemporal regression, built environment factor
IntroductionIn urban road network, the recurrent or current congestion of a certain road segment may largely impact the local network and reduce travel efficiency. Consequently, it is important to identify the Compared with mobile phone data, floating car data, cargo transport vehicle record and navigation system, taxi GPS trace data is one of the easiest available sources for accurate travel route and travel time records for a wider area with more road details. Data mining based on taxi trip can be traced back to the 1970s (Goddard, 1970), which has been applied to a wide range of studies, mainly including activity-based and infrastructure-based fields. The activity-based studies mostly focus on driver behavior, supply-demand pattern, and traffic state analysis, while the infrastructure-based studies mainly focused on lanes channelization (Tang, Yang, Kan, & Li, 2015) and signal-timing estimation (Yu & Lu, 2016).From driver behavior perspective, Zhang, Qiu, Duan, Du, and Lu (2015) proposed a space-time visualization method to demonstrate taxi daily trajectories by GIS-T to recognize working time, operating range, and residence location without time division. Qing, Parfenov, and Kim (2015) compared direct extracted datas like travel distance, speed, demand, and supply mismatch of taxi trip between fair weather and extreme storm using Manhattan GPS data, and discovered the reduction in trip distance and supply of drives during the extreme storm. Meanwhile, Hwang, Wu and Jian (2006) used structural equation modeling technique...
