Parking Occupancy Forecasting: A Methodological Analysis

In urban areas, parking presents a persistent challenge, with limited spots and increasing traffic congestion topping the list of concerns. The scarcity of available parking frustrates people and contributes to traffic jams. The study aims to create a predictive model capable of accurately predicting parking occupancy by leveraging historical data. This comprehensive study delves into parking space occupancy predictions for a college campus garage, employing various models. The research utilizes a dataset from 2022 through March 2023 and employs random forest, decision tree, linear regression, and support vector regression models to compare their effectiveness using evaluation measures. Random forest demonstrated superior performance compared to decision tree, SVR, and linear regression models in terms of accuracy, as evidenced by lower MAE values. Moreover, random forest exhibited enhanced classification performance, and a higher R2 value indicated a more robust correlation between predicted and actual occupancy values. Despite slightly higher errors than the random forest, the decision tree model's interpretability and ability to capture parking occupancy's underlying factors make it a valuable tool for understanding parking dynamics. In contrast, SVR exhibited the poorest performance among the four models, with a high MAE and a negative R2 value. The results of this study will assist in enhancing parking management systems and contribute to the development of efficient parking solutions.