Nonlinear effects of changes in the built environment and life events on mode choice: A longitudinal analysis
利用荷兰五波面板数据，用机器学习模型分析建成环境变化和生活事件对出行方式选择的动态非线性影响，发现生活事件和人口特征影响最大，建成环境变化影响有限。
Many studies have investigated the relationships between the built environment attributes, life events, and travel mode choice. However, few have explored their dynamic, nonlinear relationships in a holistic framework. Using five waves of the Netherlands mobility panel data, this study applies a light gradient boosting machine model to examine the nonlinear effects of socio-demographics, changes in the built environment, and life events on evolving mode choice. We find that socio-demographics and life events have the dominant relative importance for predicting four latent evolving mode choice patterns. Results show that education, gender, employment status, change in working hours, and starting a new job are among the determinants with the highest relative importance. Most life events show nonlinear effects and a few exhibit deviated short- and long-term effects on the modal shifts between car and green modes. Changes in built environment attributes have nonlinear associations with evolving mode choice, but the impacts are rather limited. These findings offer policy implications and planning guidelines for promoting the modal shift from car to green modes toward sustainable mobility transition.