Real-Time Cloud-based Data Analysis Using Machine Learning for Smart Parking

The fourth industrial revolution has led to the emergence of smart cities and intelligent transportation systems that rely heavily on large-scale data-driven models. One of the most popular applications of these models is smart parking, which has gained significant attention from both researchers and industry professionals. However, different scenarios and environments require different approaches due to the various factors that can impact people's schedules and behavior. This paper aims to investigate traffic assignments based on parking prediction for a mid-size university environment with multiple parking garages around the campus and various types of permits in each garage. The investigation involves utilizing multiple machine learning (ML) techniques to predict parking garage correlation. To achieve this, we collected data from different parking garages and used some of this data to train our ML algorithm. The machine learning models developed in this study demonstrated significant predictive accuracy. Our baseline model achieved an RMSE of 33.15, whereas our enhanced model incorporating features like availability ratio and other temporal features reduced the RMSE to 0.0864, with near-perfect prediction accuracy (MAPE < 0.005%). These results underscore the effectiveness of feature engineering and advanced modeling in addressing the complexities of parking availability prediction.