Predicting the spatiotemporal legality of on-street parking using open data and machine learning

Searching for a parking spot in metropolitan areas is a great challenge, especially in highly populated areas such as downtown districts and job centres. On-street parking is often a cost-effective choice compared to parking facilities such as garages and parking lots. However, limited space and complex parking regulation rules make the search process of on-street legal parking very difficult. To this end, we propose a data-driven framework for understanding and predicting the spatiotemporal legality of on-street parking using the NYC parking tickets open data, points of interest (POI) data and human mobility data. Four popular types of spatial analysis units (i.e. point, street, census tract, and grid) are used to examine the effects of spatial scale in machine learning predictive models. The results show that random forest works the best with the minimum root-mean-square-error (RMSE) for predicting ticket counts and with the highest accuracy scores for spatiotemporal legality classification across all four spatial analysis scales. Moreover, several prominent categories of places such as those with retail stores, health-care services, accommodation and food services are positively associated with the number of parking violation tickets.