Analyzing Spatiotemporal Congestion Pattern on Urban Roads Based on Taxi GPS Data

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 the 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 24-h congestion pattern of road segments in an urban area. The spatial autoregressive moving average model (SARMA) was then introduced to analyze the output from the clustering analysis to establish the relationship between the built environment and the 24-h congestion pattern. The road segments were classified into four congestion levels. The regression explained 12 traffic-related factors and land use factors’ impact on road congestion pattern. The continuous congestion was found to mainly occur in the city center. Factors, such as road type, bus station in the vicinity, ramp nearby, commercial land use have large impacts on congestion formation. In combination with quantitative spatial regression, the proposed Fuzzy C-means clustering approach was employed to develop an overall evaluation process, which could be applied generally to assist the assessment of spatial–temporal levels of road service from the congestion perspective.