Using tracking data to develop an explanatory model and a prediction model for parking search durations
Consortium partners: Fluxguide Ausstellungssysteme GmbH (Viena), Bliq (Berlin)
Duration: July 2020 until June 2023
Funded by: Federal Ministry of Digital Affairs and Transport
Project Lead and Contact: Prof. Dr. Tobias Hagen
When calculating journey durations with navigation apps, the time needed to find a parking space is neglected. Therefore, the attractiveness of private car use compared to other means of transport appears higher than it actually is. Implementation of parking search duration in navigation apps could reduce unnecessary parking search traffic. This would be associated with relief in emissions, traffic volume, and travel time. Moreover, previous studies show that parking search time (cruising for parking) has not been convincingly measured yet.
This research project closes a research gap by precisely measuring parking search based on collecting data via app developed for this purpose. In this way, average parking search durations can be determined according to district types and time. One research goal of the project start2park is developing a model to explain parking search to identify public traffic-planning options. Moreover, a prediction model is developed in order to implement real-time forecasts of individual parking search durations in navigation apps.
The start2park-app is developed in collaboration with our practice partner Fluxguide and provides a state-of-the-art mobile interface that is fine-tuned for ease of use while driving. The app is available in app stores for both iOS and android devices. App-based data will be combined with big data by our practice partner Bliq. Using data mining and statistical analysis, cruising for parking will be explained by possible determinants, e.g., traffic density, date, and time. Using machine learning algorithms, a parking search time prediction model will be developed.
Prediction Model:
- The machine learning model to recognize parking searches in arbitrary floating car data is now available for free on GitHub and you can integrate it into your program code: https://github.com/ReLUT/parking-search-prediction
- On the following page, you can test the model by uploading individual trips and viewing the identified parking search: http://start2park.eu/
Further information about start2park project can be seen in these videos:
