Research on the influencing factors of urban on-street parking demand based on ensemble learning and SHAP

Previous studies have often focused on the influence of the built environment on urban parking behavior and neglected the role of socioeconomic attributes. Based on the data on on-street parking in a city, this paper characterizes the urban onstreet parking demand in terms of the parking space utilization rate of on-street parking lots while establishing a model of the influence of socioeconomic attributes and the built environment on the urban on-street parking demand by using the ensemble learning method. Besides, the model is interpreted based on the SHAP theory to reveal the influence mechanism of the built environment and socioeconomic attributes on urban on-street parking demand. The results of the research show that socioeconomic attributes are more important than the built environment in terms of the degree of influence on parking space utilization, with the economic attributes of the parcel (GDP) ranked first and the price charged for parking lots ranked second. GDP, population density, and commuting density have a positive effect on the parking space utilization rate, while the price charged for parking has a negative effect on the parking space utilization rate. Among them, the effect of GDP is almost linear, while the effect of population density, price charged, and CBD proximity are nonlinear. For example, when CBD proximity increases from 0km to 4km, its effect on parking space utilization gradually turns from negative to positive. And when it is higher than 4km, the effect turns from positive to negative.