Predictive Analytics for Parking

Team: SDOT | Google, SoftServe, Expedia | Innovative Data Acquisitions

How might we create a model that predicts street parking occupancy to set rates in Fall 2020?

Problem: During normal times, lack of payment makes it challenging to understand actual counts of vehicles parked. The project team conceived of a statistical model that could predict street parking based on historical and current transactions as well as ground truth data, to better understand demand for parking and help residents find parking. When COVID-19 hit, traffic patterns changed significantly and paid parking was suspended, cutting off new data that would be needed to validate the model. The problem-space shifted from moderating demand for street parking to setting demand-based paid parking rates that appropriately reflect demand and supply of parking during a pandemic.

Solution: SDOT released one of the largest open data sets of parking data in the country with 1.6 billion transactions. Developed interactive data dashboard with maps and charts by time of day, location and year. Dashboard offers deep statistical analysis, modified data collection plan and supports the ability to set parking rates during COVID-19 recovery.

With paid parking system re-established in early July at minimum rates, the Curbside Management team is using the data model to determine where to adjust rates this fall where parking is overly full and not providing appropriate customer turnover and access.

Contact: Mary Catherine Snyder, SDOT, marycatherine.snyder@seattle.gov