Using the traffic heterogeneity of Chinese toll highway networks for hierarchical clustering

The spatial clustering of highway traffic is of great interest to researchers and policy makers. In this paper, instead of using the microscopic traffic parameters in the traditional clustering methods, we introduce a new heterogeneity index clustering the sections of a highway based on differences in the content, a.k.a. “Heterogeneity”, in their flow, which can be used as a universal guideline for network spatial clustering. Using real-world toll station origin to destination (O-D) data in three highway networks of China, we evaluate the stability of the traffic heterogeneity and verify the strong correlation between the traffic heterogeneity and the traffic variation. A case study on the hierarchical clustering for these highway roads was carried out, and we evaluate the clustering performances and show that the heterogeneity is a better partitioning criterion than other conventional traffic indices.

I. Introduction

Understanding the spatiotemporal behaviors of the traffic is a fundamental requirement of the clustering studies in traffic data mining. Existing clustering studies focus on the microscopic traffic behaviors, such as traffic density, or congestion state as a partitioning guideline. In a enclosed highway system, the traffic can only be exchanged at a toll station, and thus the dynamic variations in upstream and downstream components of a traffic stream are fully captured by the origin and destination (O-D) data at the toll station. This information has the potential to reveal an important spatiotemporal feature that can help researchers analyze and group the highway sections under the same traffic influence.