Long-Term Forecasting of Off-Street Parking Occupancy for Smart Cities
Recent developments in the field of parking can be enhanced with smart city alternatives. One of these alternatives is the monitoring of parking sensor data. In this paper, this data is used to propose a Decision Support System (DSS) that supports the decision-making of the municipality of Amsterdam on parking. The DSS provides insight into the six months ahead parking occupancy for 57 off-street parking locations in Amsterdam. An effect analysis has been conducted into factors that influence the off-street parking occupancy, and five forecast models are compared to predict the parking occupancy. For the effect analysis, weather and event variables are highlighted. It is observed that the most influential factors on parking occupancy are sunshine, temperature, relative humidity and event factor ’match’, that indicates whether or not a soccer match is taking place. The forecasting algorithms compared are Seasonal Naive Model as a benchmark approach, Box-Jenkins Seasonal Autoregressive Integrated Moving Average with and without exogenous regressors (SARIMAX and SARIMA, respectively), exponential smoothing models, and Long Short-Term Memory neural network. Based on the effect analysis study, the exogenous regressors of the SARIMAX model are included per parking location. This model also outperforms the other algorithms according to the lowest Root Mean Squared Error. Especially the event factor is important for the parking occupancy forecasts. Future studies can focus on the addition of more event variables, the extension into an online model based on real-time parking sensor data and the effect analysis on changes, such as public transit networks on parking.
|, , , ,|
Fokker, E.S, Koch, T, & Dugundji, E.R. (2021). Long-Term Forecasting of Off-Street Parking Occupancy for Smart Cities. In Proceedings of the Transportation Research Board 100th Annual Meeting.