Comparative Evaluation of Probe-Based Travel Time Prediction Techniques Under Varying Traffic Conditions

Accurate short-term travel time forecasting is one of the most important requirements of the development of effective and reliable traffic management and traveler information systems. In this paper, probe vehicle data obtained from an AVI system is used to develop and test the effectiveness of various travel time prediction methods under different traffic conditions. Most of the previous research tested the performance of time series models using data from 'normal' traffic conditions mainly due to the reason of limited data availability. However, the essential challenge is to predict travel time under rapidly changing traffic conditions. In this study, probe vehicle data is directly collected from an electronic toll collection system of more than 140 miles long. This novel use of existing data obviates the need for additional infrastructure investment for collecting probe vehicle data. Moreover, large and diverse probe vehicle data used in this project eliminates some of the problems of past research projects caused due to lack of sufficient real probe vehicle data under different but more importantly time-varying traffic conditions. This paper compares the prediction performance of different prediction methods; a parametric method, ARIMA, and two non-parametric methods, namely support vector regression (SVR) and neural networks (NN) using probe vehicle data for different traffic conditions. In addition to single ARIMA model, a multivariate ARIMA model, which incorporates an indicator variable that signals the unexpected conditions (e.g. incident, inclement weather) is employed. Then, performance of all three models under both typical and atypical traffic conditions are analyzed and strengths and weaknesses of each model for time-varying traffic conditions are discussed.

• Supplemental Notes:
  • This paper was sponsored by TRB committee AHB20 Freeway Operations
• Corporate Authors:

  Transportation Research Board

  500 Fifth Street, NW
  Washington, DC  United States  20001
• Authors:
  • Yildirimoglu, Mehmet
  • Ozbay, Kaan
• Conference:
  • Transportation Research Board 91st Annual Meeting
  • Location: Washington DC, United States
  • Date: 2012-1-22 to 2012-1-26
• Date: 2012

Language

• English

Media Info

• Media Type: Digital/other
• Features: Figures; References; Tables;
• Pagination: 20p
• Monograph Title: TRB 91st Annual Meeting Compendium of Papers DVD

Subject/Index Terms

• TRT Terms: Forecasting; Neural networks; Parametric analysis; Probe vehicles; Travel time
• Uncontrolled Terms: Support vector regression
• Subject Areas: Data and Information Technology; Highways; Operations and Traffic Management; I71: Traffic Theory; I73: Traffic Control;

Filing Info

• Accession Number: 01371104
• Record Type: Publication
• Report/Paper Numbers: 12-3677
• Files: TRIS, TRB
• Created Date: May 25 2012 9:31AM