In this second installment of the TMIP AI Webinar Series, we will explore examples of how machine learning methods have already been incorporated into adopted travel models in recent years to improve the prediction of trip (or tour) generation. The first presentation will share how boosting was applied to improve the generation of non-home-based (NHB) trips from home-based trips by incorporating accessibility. This was the first known example of machine learning’s use for practical travel forecasting in 2017 as part of the Tennessee statewide model. Since then, the approach has been included in over a dozen models in use in at least seven states.
The second presentation will showcase how decision trees improved the prediction of home-based trips (or equivalently, tours) in the regional model for North Carolina’s Research Triangle (of Raleigh, Durham, and Chapel Hill). This was the first use of decision trees in an adopted model in 2021, and its success has inspired the use of the technique in several additional models since.
Hadi Sadr, PhD, is a Senior Manager at Iteris with over 15 years of experience specializing in travel demand modeling, including model development, calibration, and validation. He has proven expertise in regional and statewide modeling, integrating big data sources (StreetLight, INRIX, ATRI), and delivering investment-grade traffic and revenue forecasts. His experience also includes transit forecasting, toll modeling, and emerging mobility solutions such as connected and automated vehicles (CAVs) and transportation network companies (TNCs).
Si Shi, is a Research Scholar at the Institute for Transportation Research and Education at North Carolina State University where she is the lead modeler for the Triangle Regional Model in the Research Triangle Region of North Carolina. In the past 10 years she has played with various travel models and enjoyed using data and models to inform decision making. She has a master’s degree from UNC Chapel Hill in City and Regional Planning.