Identifying Relevant Predictors and Spatial Relations of Traffic Parameters for Networkwide Short-Term Traffic Prediction

With the aim of improving the predictive ability of short term traffic prediction models, this paper proposes statistical approaches to identify spatial relationships among road links in an urban road network. Specifically, these approaches provide systematic methods of selecting predictors for a short-term traffic prediction model for a given road link. The three methods are the Granger causality model, the Elastic net regularization model, and a combination of the two defined Granger-Elastic net. The three methods are applied to the urban road network of Brisbane, Australia, that was selected as the case-study for application purposes. The analysis used one year traffic flow and speed data collected at 30-minute intervals by sensors located on the selected road network. The efficiency of the proposed methods for the relevant selection of predictors was evaluated in terms of prediction accuracy of a multiple linear regression method that was applied as a short term traffic forecast model. For a given target link, the relevant predictors obtained by the Granger causality, the Elastic net and the Granger-Elastic net models were used separately to build short term traffic prediction model. The results show that Granger-Elastic net can significantly computational costs by removing irrelevant predictors. In addition, relevant predictors obtained from Granger-Elastic net can accurately forecast traffic flow and speed in the selected road network. This study also identifies the important predictors in the road network that can be used for developing traffic prediction model when the target location or a number of road links has malfunctioning detectors.

• Supplemental Notes:
  • This paper was sponsored by TRB committee ADB40 Standing Committee on Transportation Demand Forecasting.
• Authors:
  • Hasan, Md Mahmud
  • Kim, Jiwon
  • Prato, Carlo
• Conference:
  • Transportation Research Board 97th Annual Meeting
  • Location: Washington DC, United States
  • Date: 2018-1-7 to 2018-1-11
• Date: 2018

Language

• English

Media Info

• Media Type: Digital/other
• Features: References; Tables;
• Pagination: 6p

Subject/Index Terms

• TRT Terms: Linear regression analysis; Statistical analysis; Traffic distribution; Traffic estimation
• Geographic Terms: Brisbane (Australia)
• Subject Areas: Highways; Operations and Traffic Management; Planning and Forecasting;

Filing Info

• Accession Number: 01660980
• Record Type: Publication
• Report/Paper Numbers: 18-06378
• Files: TRIS, TRB, ATRI
• Created Date: Feb 22 2018 9:19AM