Short-Term Traffic Prediction Under Normal and Abnormal Traffic Conditions on Urban Roads

Short-term traffic prediction can support proactive traffic control in Intelligent Transportation Systems (ITS) to help traffic network managers anticipate and mitigate network problem in advance. In previous work on this topic, three models with increasing information in explanatory variables were developed and tested for 15-min ahead traffic prediction concerned with normal and abnormal traffic conditions using a dataset from Inductive Loop Detectors (ILDs) in central London. In this paper, the k-Nearest Neighbour (kNN) and Support Vector Regression (SVR) algorithms were used as machine learning tools for implementing the models with an objective to compare these machine learning tools using the model structures used earlier. The prediction accuracy of models implemented using the kNN and SVR methods is evaluated for normal traffic conditions and incident conditions using data from central London. This study shows that the kNN and SVR methods have the similar prediction accuracy in normal, non-incident traffic conditions. However, the kNN method outperforms the SVR approach under abnormal, incident traffic conditions. In addition, of the three different model mechanisms used, the structure of error feedback improved the prediction accuracy of the kNN based model under non-recurring abnormal traffic conditions, supporting the results from earlier studies.

  • Supplemental Notes:
    • This paper was sponsored by TRB committee AHB15 Intelligent Transportation Systems
  • Corporate Authors:

    Transportation Research Board

    500 Fifth Street, NW
    Washington, DC  United States  20001
  • Authors:
    • Guo, Fangce
    • Krishnan, Rajesh
    • Polak, John W
  • Conference:
  • Date: 2012


  • English

Media Info

  • Media Type: Digital/other
  • Features: Figures; References; Tables;
  • Pagination: 17p
  • Monograph Title: TRB 91st Annual Meeting Compendium of Papers DVD

Subject/Index Terms

Filing Info

  • Accession Number: 01366524
  • Record Type: Publication
  • Report/Paper Numbers: 12-1627
  • Files: TRIS, TRB
  • Created Date: Mar 29 2012 7:14AM