A novel integrated machine learning and inferential modeling approach to explore non-linear effects of built environment on travel: a three-wave repeated cross-sectional study

Abstract

This study investigates the dynamic effects of the built environment on travel in Austin, Texas, over a 20-year period. Using three waves of household travel surveys from 1997, 2006, and 2017, the research employs a repeated cross-sectional approach to address the limitations of traditional longitudinal and cross-sectional studies, and to more accurately estimate effect sizes. Methodologically, we introduce a novel integration of machine learning and inferential modeling to uncover non-linear relationships and threshold effects of the built environment characteristics on travel. Using Gradient Boosted Decision Trees (GBDT) and Partial Dependence Plots (PDPs), we first identify optimal threshold points in the relationships, which are then incorporated into piecewise multilevel models.

Findings from the study reveal that the built environment serves as a sustainable tool for managing travel in the long term, contributing 50 % or more to the total feature importance in predicting individual travel—surpassing the combined effects of personal and household characteristics. Improved transit accessibility, enhanced local and regional accessibility, higher population and employment densities, and greater diversity are all associated with significant reductions in travel—particularly within their identified thresholds—though the magnitude of their influence varies across time periods and shows diminishing marginal returns.

These findings highlight the potential of smart growth policies—such as expanding transit accessibility, promoting high-density and mixed-use development, and discouraging single-use development and peripheral sprawl—as effective strategies to reduce car dependency and manage travel demand. Moreover, the study demonstrates that the proposed integrated approach can effectively capture complex non-linear effects while enhancing flexibility and interpretability, reducing researcher bias, and enabling statistical inference—ultimately providing more robust and policy-relevant insights.