Developing a pedestrian destination choice model using the stratified importance sampling method

Elmira Berjisian

doi:10.1016/j.jtrangeo.2019.04.009

2019-05-01

PRE

Elmira Berjisian

Meeghat Habibian *

DOI：10.1016/j.jtrangeo.2019.04.009

原文链接

原文求助

摘要

En 中文

Pedestrian destination choice models are among the issues that have not been adequately addressed by transportation researchers. Several issues should be considered in order to develop a pedestrian destination choice model, including the method used to generate destination choice sets as well as the specification of different environmental variables that influence walking. Although some studies use built environment variables to elaborate on pedestrian destination choice models, almost all of them neglect the importance of the choice set generation method and rely merely on the random sampling method. This study illustrates the extent to which a choice set generation method can influence the parameter estimates of a pedestrian destination choice model. Here, in addition to the random sampling method, the method of stratified importance sampling is addressed to develop a pedestrian destination choice model for the city of Rasht, Iran. In addition, an exhaustive array of built environment variables that are known to have a significant effect on walking is employed. The results suggest that using stratified random sampling can improve the destination choice model in both goodness of fit and percent correct of the prediction of pedestrians' destination. The relative importance of environmental variables with respect to altering the pedestrians' destination choices is demonstrated through their elasticities' values. Two scenarios, which aim to improve the walking environment by enhancing network connectivity, are examined. As a result, the pedestrian catchment area is suggested to be the most effective variable of network connectivity. The results of these scenarios can assist policy makers to identify zones that deserve more attention.

Keyword:

Walking

Destination choice model

Choice set

Stratified importance sampling

Rasht