Short-Term Traffic Flow Prediction Based on the IMM-BP-UKF Model

This study aims to obtain satisfactory short-term traffic flow forecasting results in various traffic flow modes. Clustering analysis of traffic flow data is carried out using the Kohonen neural network. Then, different BP neural networks are trained using different classified traffic flow data. Different neural network models are combined with UKF to form multiple traffic estimators to realize the estimation function. Finally, the interactive method is used to fuse the prediction results of each estimator, and comprehensive traffic flow forecasting results are obtained. With the simulation example, a number of single estimators are obtained, and a joint estimator based on this method is constructed. The section flow of one given road is predicted to verify the performance of the estimator. Results show that the joint estimator has a higher prediction accuracy than the single estimator and has adaptive characteristics when traffic flow characteristics change.

  • Record URL:
  • Supplemental Notes:
    • © 2019 American Society of Civil Engineers.
  • Authors:
    • Zhou, Xing-yu
    • Li, Hong-mei
    • Zheng, Wei-Hao
    • Tang, Zhi-hui
    • Yang, Li-jun
  • Publication Date: 2019-6


  • English

Media Info

Subject/Index Terms

Filing Info

  • Accession Number: 01720079
  • Record Type: Publication
  • Files: TRIS, ASCE
  • Created Date: Aug 2 2019 3:04PM