建成环境非线性作用下汽车客运站最高聚集人数模型

Model of the maximum number of people gathered in a bus terminal considering the non-linear effect of the built environment

为探求汽车客运站最高聚集人数与站域周边范围建成环境的非线性作用机制,以重庆市全域汽车客运站为对象,构建以最高聚集人数为因变量,由可达性、路网复杂度、公共交通支撑、多样性4个方面组成的多元建成环境特征为自变量的梯度提升决策树模型,辅助构建随机森林和最小二乘估计模型验证算法效度。结果表明:GBDT模型在训练集和测试集的表现均优于RF和OLS模型,验证集拟合优度为0.80;变量贡献度方面,土地利用混合度对影响最高聚集人数的建成环境变量的贡献为23%,其次为公交站密度(贡献为20%)和至市中心距离(贡献为17%);非线性作用机理方面,建成环境可达性相关指标总体呈负相关,阈值效应明显,路网复杂度相关指标峰值效应较明显,公共交通支撑相关指标整体呈正相关,多样性相关指标峰值效应明显。

To explore the nonlinear mechanism between the maximum number of people gathered in the bus terminal and the built environment around the terminal,a gradient boosting decision tree(GBDT)model with the maximum number of people gathered as the dependent variable and the multiple built environment characteristics composing of accessibility,road network complexity,public transportation support,and diversity as the independent variables is constructed with assistance of random forest(RF)and least squares estimation model(OLS)for algorithm validation.The results show that the GBDT model outperform the RF and OLS models in the training and test sets,with a goodness-of-fit of 0.80 in the validation set;in terms of variable contributions,land use mix is the built environment variable that influence the highest number of congregations with a contribution of 23%,followed by bus stop density(20%)and distance from the city center(17%);in terms of nonlinear mechanism of action and built environment accessibility-related,the indicators related to accessibility of built environment are negatively correlated and the threshold effect is obvious,the peak effect of indicators related to road network complexity is obvious,and that of indicators related to diversity is also obvious,the indicators related to public transportation support are positively correlated.