Mixed parking demand assignment in hub parking lots based on regression modeling

Abstract

To adapt to the unique demand-supply features of accessory parking lots at passenger transport hubs, a mixed parking demand assignment method based on regression modeling is proposed. First, an optimal model aiming to minimize total time expenditure is constructed. It incorporates parking search time, walking time, and departure time, focusing on short-term parking features. Then, the information dimensions that the parking lot can obtain are evaluated, and three assignment strategies based on three types of regression models—linear regression (LR), extreme gradient boosting (XGBoost), and multilayer perceptron (MLP)—are proposed. A parking process simulation model is built using the traffic simulation package SUMO to facilitate data collection, model training, and case studies. Finally, the performance of the three strategies is compared, revealing that the XGBoost-based strategy performs the best in case parking lots, which reduces time expenditure by 29.3% and 37.2%, respectively, compared with the MLP-based strategy and LR-based strategy. This method offers diverse options for practical parking management.

Keywords

parking areas assignment / hub parking lot / regression-based modeling / extreme gradient boosting (XGBoost)

Funding

National Natural Science Foundation of China(52302388)

Natural Science Foundation of Jiangsu Province(BK20230853)