Spatio-Temporal Autocorrelation-Based Clustering Analysis for Traffic Condition: A Case Study of Road Network in Beijing

Abstract

Traffic congestion is an increasingly serious problem worldwide. In the last decade, many cities have paid great efforts to establish Intelligent Transportation Systems (ITS), and a large amount of spatio-temporal data from traffic monitoring system is also accumulated. However, with the devices and facilities of ITS getting completed, effectiveness of ITS practices is always restricted by traffic information fusion and exaction technique. Traffic condition-determining is a crucial issue for Advanced Traffic Management Systems, on which many researchers have done profound studies. The existing studies are mostly focused on traffic condition recognition at a certain road and time point; while in practice, it’s more meaningful how different kinds of traffic condition are correlated and distributed in space-time. Therefore, in this research we present an improved spatio-temporal Moran scatterplot (STMS), by which traffic conditions are pre-classified into four types: homogenous uncongested traffic, heterogeneous uncongested traffic, homogenous congested traffic and heterogeneous congested traffic. Then at the basis of STMS, a novel spatio-temporal clustering method combining pre-classification of traffic condition is proposed. Finally, the feasibility and effectiveness of the clustering methodology are demonstrated by case studies of Beijing. Result shows that the proposed clustering method can not only effectively reveal the relation of traffic demand to road network facilities, but also recognize the road sections where congestion originates or gets alleviated in the network, which provides foundations for traffic managers to alleviate congestion and improve urban transport services.