Research on Traffic Congestion Patterns Based on the Traffic Performance Index: The Case of Beijing

Congestion has become a global problem with urbanization and motorization. Research on the changing laws and typical traffic congestion patterns is essential for developing effective traffic management strategies. This paper aims to create a clustering method to classify congestion patterns based on the traffic performance index (TPI) collected from April to June 2017 in Beijing. New techniques are proposed using the k-means clustering algorithm by analyzing the traditional clustering method in a current study, which determines the number of clustering categories based on silhouette. The coefficient of variation and the coverage ratio evaluates the results of congestion patterns. The new method using TPIs weighted by the coefficient of variation can effectively classify traffic congestion into typical patterns. The case study shows that TPI categories are highly correlated between 2016 and 2017. Therefore, this method is helpful for traffic management, making decisions according to different congestion patterns.