��������

�������� transportation professor Guang Tian has been awarded a grant to study traffic patterns and determine the best model for predicting how many miles motorists travel in a given area and time frame.

Tian’s research, which will be conducted over the next year, is funded by the Louisiana Transportation Research Center. It seeks to provide transportation engineers, planners and policymakers with the right model of predicting vehicle miles traveled (VMT) in order to manage traffic and congestion, plan future investments, control emissions and address other issues, Tian said.

Vehicle miles traveled is a transportation planning tool that’s used to measure the amount of travel for all vehicles in a geographic region over a designated time and is used to evaluate the performance of transportation systems.

“In the last couple of decades, there has been a paradigm shift of transportation performance measurement from how fast vehicles move to how well people’s travel needs are met,” Tian said. “And from speed to mobility, accessibility, sustainability and livability, and from level of service to VMT.”

VMT is the key indicator of the transportation system and marks both the positive (economic growth, personal mobility) and negative externalities (congestion, crash, emissions) of automobile use, Tian said.

“VMT or VMT per capita has been used by federal, regional and local agencies to evaluate the performance of their transportation systems,” Tian said.

Tian will use a household travel survey database that includes more than 1 million trips generated by 100,000 households across 36 metro areas in the U.S. to model and predict VMT. ��

His research will also evaluate and compare the prediction performance of statistical models versus machine learning models to determine which is the right model for predicting vehicle miles traveled.

“The prediction performance of machine learning on VMT has not been tested and systemically evaluated by a large multi-regional database and compared against traditional statistical models,” Tian said. “This proposed research aims to explore the application of machine learning in predicting VMT and to compare its prediction power with traditional statistical methods by using a large database.”

��

��