A Study on the Heterogeneity of Travel Purposes in Pedestrian Route Choice: Based on a Hierarchical Bayesian Path Size Logit Model

Abstract

In high-density urban areas, pedestrian route choice behavior is influenced differently by the attributes of the objective environment, depending on the purpose of the trip. To explore this issue, this paper first constructs the traditional Path Size Logit (PSL) model to perform baseline estimation of the effects of objective attributes such as path distance, intersections, number of lanes, greenery, and commercial facilities. It then introduces a hierarchical Bayesian framework to build the Hierarchical Bayesian Path Size Logit (HB-PSL) model, using the No-U-Turn Sampler (NUTS) for posterior sampling to quantify parameter uncertainty and capture inter-group heterogeneity. The model achieves inter-group information sharing through a hierarchical prior structure and uses Automatic Differentiation Variational Inference (ADVI) to provide rapid approximate estimation. An empirical analysis shows that different importance is given to objective environmental attributes for different travel purposes. Furthermore, this paper proposes the “Equivalent Distance (ED)” index, which transforms the preference of different groups for different environmental attributes into an actionable spatial length. For shoppers, the greening level increases by one level for every 100 m, which is equivalent to shortening the path by about 31.31 m; commercial facilities increase by one for every 100 m, which is equivalent to shortening the path by about 76.63 m. The results provide behavioral support for the formulation of differentiated walking strategies in high-density urban areas: priority should be given to strengthening the green coverage and commercial facility layout in commercial blocks to synergistically improve walking efficiency, safety, and comfort.