Module for integrating parking hubs with the parking lot occupancy forecasting system

The growing number of vehicles in cities creates complex challenges for parking management systems that require effective tools for predicting parking congestion. The purpose of this study was to develop and implement an integrated module for predicting parking space congestion in real time. To achieve this goal, a hybrid approach to data processing was applied, combining machine learning methods with time series analysis and spatial dynamics, and integration with modern software technologies. The results of experimental testing showed an increase in the accuracy of forecasts of parking space congestion by 20-25% compared to conventional methods, which significantly contributed to the rapid response to dynamic changes in the urban environment. By automating real-time data collection, cleaning, and aggregation, information update delays have been reduced by 10-12%, providing a more up-to-date and reliable analytical framework for management decisions. However, the increased accuracy of forecasts and prompt access to updated data helped to increase the efficiency of using parking spaces by 15-20%, optimising the distribution of traffic flows, and reducing congestion. The implementation included the use of Java and Spring Boot 3 for backend logic, AWS S3 for cloud storage, PostgreSQL as the main database, and Python algorithms using NumPy, Pandas, and python-dateutil for machine learning. Statistics, trends, and forecasts were visualised using React, which allowed users to get interactive access to results and make informed decisions. In addition, the module is easily scalable, adapts to different types of infrastructure, and can be successfully integrated into existing parking management systems. The practical significance of the development is to improve the quality of urban life by reducing congestion, reducing the environmental burden and rationalising the use of urban transport infrastructure