Short-term traffic flows in urban road networks have significant random and non-linear features. Researches on the spatiotemporal correlations of short-term traffic flows can demonstrate the entire evolution process of traffic congestion life cycles, thus helping traffic managers make traffic dispatches or interventions in advance for the traffic flows in the entire traffic road networks, so as to improve the traffic capacity of road networks. Firstly, the existing Pearson correlation analysis method is improved in this research first by separating floating car data and extracting the trends and detail vectors of road traffic flows, and introduction of the generalized least squares method makes Pearson method closer to the evolution of road traffic conditions. Secondly, based on the characteristics of traffic state classification, the Kendall coefficient is introduced to study the evolution of road traffic flow states. And finally, a high-correlation path search method based on fusion coefficients is proposed to provide method supports for traffic congestion prediction.

CICTP