High-Dimension Traffic Data Imputation Based on a Square Norm

Traffic data is often missing or incomplete. Recently, it has been constructed into a tensor model and its missing data can be completed from a subset of its observed entries to fully express the multi-modal properties of traffic data. However, with the dimension of the data becoming higher, the completing speed, and accuracy usually decreases rapidly. To solve this problem, in this paper, the authors introduce a square-norm model for tensor completion, which uses the matrix completion method to accelerate the procedure of iterations, which makes it more adapted for higher-dimension tensor completion. The experimental results show that their algorithm is more suitable for those tensors with dimensions greater than three, and the accuracy can be ensured even when the missing ratio is as high as 80%.


  • English

Media Info

  • Media Type: Web
  • Features: References;
  • Pagination: pp 284-294
  • Monograph Title: CICTP 2016: Green and Multimodal Transportation and Logistics

Subject/Index Terms

Filing Info

  • Accession Number: 01606695
  • Record Type: Publication
  • ISBN: 9780784479896
  • Files: TRIS, ASCE
  • Created Date: Jun 29 2016 3:03PM