4.4 Using Machine Learning for Travel Time Prediction in the Colorado Rockies

Tuesday, 24 January 2017: 11:15 AM
310 (Washington State Convention Center )
William Petzke, NCAR, Boulder, CO; and G. Wiener, T. Brummet, S. Linden, and A. Anderson

The I70 Corridor from Denver International Airport to Vail, Colorado is a popular, year-round driving route for access to the multitude of outdoor recreational opportunities provided by the Colorado Rockies, as well as an important east to west commercial transportation artery.  Due to heavy traffic, variable speed limits, steep/winding roads, and rapidly changing weather conditions, predicting travel time along this corridor can be challenging.

The National Center for Atmospheric Research (NCAR) has recently developed a system to provide real-time travel time prediction to assist the Colorado Department of Transportation (CDOT) with managing traffic patterns and maintenance/operations of I70.  The core of this system harnesses supervised machine learning algorithms to predict travel times over the next day on the road segments that make up the corridor.  This presentation will discuss aspects of the feature selection, data cleaning, and machine learning algorithms that were utilized in the development of the system, as well as approaches that were tried but rejected.  Verification statistics will be shown and performance of the machine learning models discussed.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner