Clustering analysis of congestion patterns based on traffic congestion index: evidence in Hangzhou, China

Cluster analysis of traffic congestion patterns based on quantitative evaluation index is crucial for better understanding traffic congestion situations and assessing urban traffic to achieve the goals of congestion situation deduction and policy formulation. This study establishes three traffic congestion clustering analysis models based on traffic congestion index in Hangzhou, China: the traditional clustering indicators based on the geometric shape of the traffic index curve (Model I), the simplified clustering indicator system characterized by geometric features of traffic index time-series curves (Model II), and the clustering indicator system based on raw sequences of traffic congestion index (Model III). The effectiveness of the clustering methods is validated by calculating the intra-class coefficient of variation (CVintra), the inter-class coefficient of variation (CVinter), and the coverage rate (CR) for the three models. The result shows that the Model III performs the best, with a coverage rate as high as 93.55%, effectively identifying urban traffic congestion patterns.