基于机器学习模型的建成环境对小汽车拥有行为的影响

为分析家庭小汽车拥有行为，同时考虑居住地和工作地建成环境的影响，构建梯度提升迭代决策树(GBDT)模型；分析社会经济属性，居住地、工作地建成环境属性对小汽车拥有行为的影响程度，并基于长春市居民出行调查数据进行实证研究.结果表明：3类影响因素中，社会经济属性对小汽车拥有行为的影响最大(58.95%)；职住地建成环境属性均对家庭小汽车拥有行为具有显著影响，且居住地建成环境影响(23.77%)高于工作地建成环境(17.28%)；职住地建成环境属性中，除居住地交叉口密度，工作地到中央商务区(CBD)距离及公共交通站点密度外，其他建成环境属性对小汽车拥有行为的影响均大于5%.因此，有必要同时优化职住地的建成环境来抑制小汽车拥有量的增长.

Exploring Influences of Built Environment on Car Ownership Based on a Machine Learning Method

To analyze the car ownership behaviors, a gradient boosting decision tree (GBDT) method is employed to explore the effect sizes of residential and workplace built environments on car- ownership decisions. The empirical analysis is conducted based on the Changchun household travel survey data. The results show that the socio-economic factors contribute 58.95% to automobile ownership collectively and rank the first among the three categories of factors. The residential and workplace built environment variables are both associated with car ownership. And the residential built environment is more influential than the workplace built environment. Except for intersection density at residential locations, distance to the central business district(CBD), and bus stop density at workplace locations, all built environment variables have relative importance more than 5%. Therefore, it is of great importance for urban planners and policy makers to optimize the urban built environment to mitigate the increase of car ownership.