Details:

Creators:
Zhao, Mo ; Appiah, Justice
Zhao, Mo ; Appiah, Justice Less -
Corporate Creators:
Virginia Transportation Research Council (VTRC)
Corporate Contributors:
Virginia. Department of Transportation ; United States. Department of Transportation. Federal Highway Administration
Virginia. Department of Transportation ; United States. Department of Transportation. Federal Highway Administration Less -
Subject/TRT Terms:
[+]
Forecasting Information Processing Interstate Highways Travel Time
Publication/ Report Number:
VTRC 23-R12
Resource Type:
Tech Report
Geographical Coverage:
Virginia ; United States
Virginia ; United States Less -
Edition:
Final Report
Corporate Publisher:
Virginia Transportation Research Council (VTRC)
Abstract:
Accurate prediction of travel time reliability measures would help state departments of transportation set performance targets and communicate the progress toward meeting those targets as required by the Moving Ahead for Progress in the 21st Century Act (MAP-21). In a recent Virginia Transportation Research Council study, Methods to Analyze and Predict Interstate Travel Time Reliability, researchers developed and tested statistical and machine learning models to analyze and predict travel time reliability on interstate highways. The generalized random forest (GRF) model showed promise in terms of data processing (no need for pre-clustering of travel times) and the relative accuracy of the results and was recommended for further evaluation by the study’s technical review panel. The current study directly adapted the previously developed GRF models to meet the requirements of MAP-21 federal target setting. In particular, the GRF approach developed using the INRIX Traffic Message Channel network for weekday peak period traffic by the prior study was successfully (1) adapted to the federally required National Performance Management Research Dataset (NPMRDS) network, and (2) expanded to cover the weekday midday and weekend daytime periods. The technical review panel was also interested in practical steps to implement the predictive models. To that end, suggested procedures for applying the new GRF models—including relevant model inputs and data preparation steps—are documented in this report. Direct application of the GRF models trained with INRIX data (2017-2018) to predict travel time reliability measures in 2009 on the NPMRDS network highlighted the need for developing new GRF models targeted to the NPMRDS network, especially when the 90th percentile travel time was predicted. Whereas the INRIX models showed mean absolute percentage errors of 37% and 51% for freeway and interchange segments, respectively, for the PM peak hours, the new GRF models (trained with 2017-2018 NPMRDS data) had relatively smaller mean absolute percentage errors of 34% for freeway segments and 38% for interchange segments depending on how work zones were characterized and how data were aggregated. Because operational improvements are often evaluated on the basis of how they improve reliability, especially on how the 90th percentile travel time is affected, the new GRF models are relevant for planning operational investments. In addition, because many of these improvements affect interchanges, the remedy of the new GRF models is essential for evaluating weaving strategies or traveler information systems that could be implemented at these locations.
Format:
PDF
Funding:
120025
Collection(s):
US Transportation Collection
Main Document Checksum:
[+]
urn:sha256:b87ba95911c42ae47a4a44bd4336d781552bc9d8303cc379cda9e2c0a7a6a1ce
Download URL:
https://rosap.ntl.bts.gov/view/dot/66451/dot_66451_DS1.pdf
File Type:

Supporting Files

• No Additional Files

More +