Anticipating changes to public transportation ridership demand is important to planning for and meeting service goals and maintaining system viability. These changes may occur in the short- or long-term; extensive academic work has focused on bettering long-term forecasting procedures while improvements to short-term forecasting techniques have not received significant academic attention. This dissertation combines traditional forecasting approaches with multivariate regression to develop a transferable short-term public transportation ridership forecasting model that incorporates fuel price as a prediction parameter. The research herein addresses 254 US transit systems from bus, light rail, heavy rail, and commuter rail modes, and uses complementary methods to account for seasonal and non-seasonal ridership fluctuations. Models were built and calibrated using monthly data from 2002 to 2007 and validated using a six-month dataset from early 2008. Using variable transformations, classical data decomposition techniques, multivariate regression, and a variety of forecasting model validation measures, this work establishes a benchmark for future research into transferable transit ridership forecasting model improvements that may aid public transportation system planners in an era when, due to fuel price concerns, global warming and green initiatives, and other impetuses, transit use is seeing a resurgence in popularity.
more about urban transportation: