Transportation Research Board Showcase Paper for Stochastics Researchers

Their article "Long-Term Forecasting of Off-Street Parking Occupancy for Smart Cities was" selected by the Transportation Demand Committee from more than 100 papers submitted in 2020.

Publication date: 04-03-2021

Transportation Research Board (TRB) has awarded CWI Stochastics researchers Elisabeth Fokker, Thomas Koch, and Elenna Dugundji the honor to present their work at a showcase of best papers at TRB’s 100th Annual Meeting. Their article Long-Term Forecasting of Off-Street Parking Occupancy for Smart Cities was selected by the Transportation Demand Committee from more than 100 papers submitted in 2020. The showcase consists of a dedicated lectern session at the centennial edition. This edition had a top record of over 19,000 registrations. The research was done as part of the long-term research partnership between CWI and the Municipality of Amsterdam.

Off-street parking occupancy forecasts

The authors focused on the prediction of occupied off-street parking spaces in Amsterdam, in order to consult the Municipality of Amsterdam to strategically develop parking supply and pricing. The goal of these developments is to reduce the number of cars in the city center, and ultimately improve the accessibility and livability of the city. The authors propose an interactive real-life operation tool that provides off-street parking predictions up until six months ahead.

The problem is solved with a Box-Jenkins Seasonal Autoregressive Integrated Moving Average with exogenous regressors (SARIMAX) model. The exogenous regressors included are event factors, such as the schedule of soccer matches and music events, and weather factors, such as the hour sum of precipitation. The event factors improved the model with 24%, and only the inclusion of weather variables reduced the model error with 8%. They have found that by including temporal patterns and external factors, the model can predict the actual values accurately in a large time span of six months ahead. Furthermore, they developed a real-time Long Short-Term Memory model to predict the parking occupancy for one hour ahead. These short-term predictions are more precise, and can be applied for policy making during events.

CWI's Stochastics Research Group