Parking Availability Prediction Using Deep Learning Approaches: Temporal Convolutional Networks (TCN) and TimesNet

Abstract

The concept of 15-minute city is gaining more attention not only as a powerful tool for decarbonization or improving public health and community-building, but also to build a coalition for a positive, forward-looking vision, and equitable future. In a traditional car-centric city, the goal of parking management is to enhance convenience for vehicles (roads, parking management, infrastructures, signs). In a 15-minute city, the goal radically shifts toward minimizing car use, and parking management becomes a tool to discourage unnecessary car trips, reclaim public space for people, and ensure equitable access for residents and essential services. Parking policies are one of the most direct levers a city can pull to change travel behavior and support the 15-minute model. The prediction of parking availability provides the intelligence to make those policies smarter and more effective. In this chapter, we compare the performance of two deep learning models for the prediction of parking availability based only on historical occupancy data.