Exploring non-linear and spatially non-stationary relationships between commuting burden and built environment correlates

The heavy commuting burden group is growing rapidly. The imbalance in commuting hours may damage traffic equality. Exploring the elements associated with commuting burden is crucial for promoting transport equity. However, few studies considered non-linear and spatially non-stationary characteristicss when exploring their relationships. In this paper, the traditional gradient boosting decision tree (GBDT) model is improved by combining it with the geographically weighted regression model, to identify nonlinear correlation and spatial nonstationarity simultaneously. The results show that there is a nonlinear correlation between the commuter burden and all the potential explanatory variables selected in this paper. And the correlation between each potential explanatory variable and commuter burden is spatially heterogeneous. The densification of public transportation and jobs can shorten the commuting time, but the convenience is diminishing. In addition, a balanced job-housing ratio, the moderate mixing distribution of different land-use types, and the development of a new urban center at an appropriate location are also conducive to easing the commuting burden. We also identify the strong non-linear correlation regions between commuting burden and built environment factors, which can provide more reference for the planning to achieve traffic equity.