A Predictive Parking Mobile Application Integrating Historical Usage and Precipitation Data

The scarcity of urban parking contributes to traffic congestion, time wastage, and environmental pollution. Existing parking management applications often provide only real-time parking availability and lack predictive capabilities for guiding drivers to available parking spots, considering, for instance, travel time and weather conditions. As its first contribution, this paper presents a predictive parking system that leverages machine learning techniques, particularly Long Short-Term Memory (LSTM), in order to predict parking availability, based on the user's needs and on historical and real-time data. It also investigates whether integrating precipitation data could improve the accuracy of parking availability predictions, an aspect that the current literature is not definitive. Evaluation, using real open datasets from Singapore, across multiple parking lots, demonstrated that incorporating precipitation provided only slight improvements in predictive accuracy. Additionally, as its second contribution, this paper presents an end-to-end cross-platform mobile application developed on top of the predictive models we propose. (Source code for the proposed application is publicly available at https://github.com/black-popcorn/SmartParkingSystem.)