The active travel (including walking and cycling) is closely related with elderly's mobility, physical and mental health, and quality of life. Hence, it is of great importance for urban and transportation planners and practitioners to examine the impacts of the built environment on elderly's active travel. Among the rich research findings, the existing research tends to focus on the built environment surrounding the living space (usually the departure place) of the elderly, while ignoring the built environment surrounding the destinations. Moreover, in prior studies, the associations between the built environment and the elderly's active travel are oftentimes assumed to be linear or log-linear. Against this backdrop, this study, taking Xiamen as an example, utilizes one of the latest machine learning method, i.e., the Extreme Gradient Boosting Decision Tree model (XGBoost) to disentangle the complex non-linear relationships between the built environment surrounding both departure places and destinations and elderly's active travel. It is found that (1) the trip distance is the most important factor that impacts elderly's propensity of the active travel; (2) the collective relative importance of the built environment variables is much higher than that of the socio-economic variables; (3) obviously, the associations between all built environment variables and elderly's active travel are non-linear and there exist "threshold" effects; (4) for some built environment variables, their impacts on elderly's active travel differ between the departure places and the destinations, while for the others, they are very similar. This study can provide a knowledge base and rich policy implications for the urban and transportation development and planning in China in the era of aging population.

built environment  /  the elderly  /  active travel