The empirical relationships between the built environment and transit use can support policy interventions for transit promotion. Limited studies have emphasized housing renters who are likely to be transit-dependent people and the nonlinear effects of built environment attributes on renters’ behavior. Using household travel data in Beijing, this study uses a decision-tree based gradient boosting machine to explore the nonlinear and threshold relationships between built environment attributes and commuting by transit. Renters are more sensitive to access to transit than owners. The collective contributions of bus stop density and distance to metro station are about 22% for renters and 14% for owners. Furthermore, most variables show non-linear effects on commuting by transit. The effects of bus stop density on renters’ commuting by transit rise sharply twice. One threshold is at a low-density level and the other is at a high-density level. Exploiting the threshold effects can produce cost-effective outcomes.
|Original language||English (US)|
|Journal||Transportation Research Part D: Transport and Environment|
|State||Published - Nov 2022|
Bibliographical noteFunding Information:
This work was supported by the Grant from Beijing Outstanding Young Scientist Program (JJWZYJH0120190003010) and the National Natural Science Foundation of China (71874010, 51878019).
© 2022 Elsevier Ltd
- Land use
- Machine learning
- Rental housing
- Threshold effect