Modeling Toll Price Impacts on Traffic Volume in New York City

New York City (NYC) is currently developing a congestion pricing policy called the Central Business District Tolling Program (CBDTP). One of the organizations that has advocated for congestion pricing is the Regional Planning Association (RPA) who also serves as the client for this capstone project. This project attempts to answer the following research questions: How can the CBDTP policy decisions be explored for the best possible options for toll prices using traffic volume modeling? How does the toll prices on NYC bridges and tunnels impact traffic volume? This capstone project utilizes mixed methods in an exploratory-sequential approach design. The principal method applied is a quantitative study with a complementary preliminary qualitative study. The qualitative methods applied are semi-structured interviews with professionals in the transportation planning sector who provided information on how planners typically approach understanding policy impacts such as the CBDTP. The quantitative research is conducted in a three part methodology: defining variables, using supervised learning modeling to analyze the correlations between historical traffic volumes and the other features analyzed, and predicting traffic volumes for each crossing based on predicted values of the independent variables using the best performing model. The dependent variable for the analysis will be traffic volume. Independent variables for the analysis include toll price, population, economic factors and temporal events and include 142 features in total that were collected and organized. 42 different Machine Learning models are tested. A Gradient Boosted Regression Trees (GBRT) model is found to predict traffic volumes with high accuracy using only toll prices and employment. This model as well as the limitations and findings of the analyses are incorporated into a report as well as a website for the client, RPA.