Data-driven traffic congestion patterns analysis: a case of Beijing

With the rapid increase of urban population and the number of motor vehicles, the traffic congestion problem is becoming more and more serious in megacities. This paper aims to identify the traffic congestion patterns and analyze their spatial–temporal variations by conducting cluster analysis for daily traffic congestion index curves. First, since the importance of sampling points in different time segments is not exactly the same, the coefficient of variation is taken to assign weight for improving K-means clustering algorithm. The improved weighted K-means clustering algorithm is proposed to identify the representative traffic congestion patterns. Second, the paired t-test method is used to analyze the spatial and temporal variations of traffic congestion patterns. Finally, case studies are conducted based on the real-life traffic congestion index data in Beijing, including over 670, 000 records covering six districts from January 1, 2017 to December 31, 2017. The results illustrate that traffic congestion patterns are both temporal dependent and spatial dependent, and the automobile license plate restriction has significant influence on the traffic congestion patterns. This study could be instructive for formulating specific traffic optimization and control schemes to alleviate congestion and promote the balance of traffic conditions.