Mining spatio-temporal patterns of congested traffic in urban areas from traffic sensor data

Road traffic condition in cities are complicated by the daily, weekly, seasonally, and weather-induced traffic demand fluctuations and the effects caused by the control of traffic signals. Therefore, it is difficult to quantitatively analyze typical traffic congestion patterns that are represented by the time and place of occurrence, the process of propagation and diminution, duration time, and many others. This study proposed a method to enumerate traffic congestion patterns from traffic sensor data based on frequent pattern mining developed in information science to understand the present situations of traffic congestion in cities. The feasibility and effectiveness of the proposed method have been evaluated through the analysis of typical congestion patterns using the traffic sensor data in Okinawa, Japan.