Revealing the built environment impacts on curbside freight parking demand using a deep generalized additive modeling framework

Abstract

The increasing intensity of urban activity has heightened conflicts between commercial vehicles curbside parking and limited road space. However, limited knowledge exists on how the urban built environment affects curbside freight parking, which restricts effective policymaking. This study investigates the spatial patterns of curbside freight parking demand using large-scale commercial vehicle trajectory data and quantifies the impacts of the built environment using an interpretable generalized additive neural network (IGANN) model. Unlike black-box models, IGANN achieves superior interpretability, reliability, and predictive accuracy. Results show that spatial heterogeneity in freight curbside parking demand is pronounced in high-density cities, where intensive development amplifies parking needs and worsens congestion. These findings highlight the importance of adopting spatially differentiated development strategies to balance freight operations with competing curb users.