PROJECT TITLE :

On-Street Parking Occupancy Inference Based on Payment Transactions

ABSTRACT:

It is essential to keep a record of both the spatial and temporal occupancy of parking spots in order to maximize the use of on-street parking resources and to devise parking policies that work. These data are typically obtained through the use of sophisticated and pricey technologies for monitoring occupancy. In addition, it is typically difficult to integrate the parking data and occupancies from other systems into bay-level occupancies. For a wide variety of analytical and practical purposes, accurate data on occupancy and payments are required. These purposes include, but are not limited to, the investigation of payment behavior, the estimation and forecasting of occupancy, and the evaluation of the efficiency and effectiveness of enforcement policies. The purpose of this study is to integrate snapshots of bay-level parking occupancy, which are captured using simple cameras, with transactions from a conventional parking payment management system. The metaheuristic optimization algorithm that is proposed for this purpose is called the metaheuristic optimization algorithm. The integrated data that were produced were utilized in the development, calibration, and validation of a method for estimating parking occupancy that solely relied on parking payment data. It is explained in detail how the proposed algorithm and modeling technique should be designed, how it should be implemented, and how it should be validated. In order to fine-tune the parameters of the data integration algorithm, logistic regression analysis was utilized. Deep learning, gradient boosting, and random forests were some of the techniques utilized in the process of developing a model of parking occupancy. The evaluation of the algorithm revealed that it had an accuracy of 76% when it came to the correct integration of data. This meant that individual bay occupancies were successfully integrated with the appropriate payment transactions. When tested with a random sample taken from the integrated data, the best occupancy estimation model demonstrated extremely high accuracy, with a R 2 value that was greater than 94% and a root mean square error (RMSE) that was 1.2 (occupied bays).