A Novel Prediction Method of Traffic Flow: Least Squares Support Vector Machines Based on Spatial Relation

Traffic situations have a great deal of spatial relationship based on the characteristics of the road network and the continuity of traffic flow. The relationship between the elements of the network can be better understood with the spatial assembling characteristics of the traffic condition, which facilitates the transportation operation, management and prediction to some extent. This paper focuses on the spatial relationship between traffic conditions of the road sections of expressways in Shanghai based on the field data collected by loop inductors. Then, the Eigen value matrix of real traffic flow is calculated and the probability distributions of these values are analyzed. With the study mentioned above, road groups (which exist in some spatial relationships) can be classified by random matrix theory. Furthermore, two training models are compared. Training data of the first model is generated from the road groups with identified spatial relationship by least square support vector machines, based on which the traffic condition is predicted. The other model functions are based on the upstream and downstream transects of the predicted road sections. This paper draws the eventual conclusion that taking spatial relation into account will definitely improve the accuracy of prediction.


  • English

Media Info

  • Media Type: Web
  • Features: References;
  • Pagination: pp 1807-1818
  • Monograph Title: CICTP 2014: Safe, Smart, and Sustainable Multimodal Transportation Systems

Subject/Index Terms

Filing Info

  • Accession Number: 01532345
  • Record Type: Publication
  • ISBN: 9780784413623
  • Files: TRIS, ASCE
  • Created Date: Jul 2 2014 3:03PM