中国城际联系距离衰减的非线性规律——基于人口流动大数据

以人口流动为表征的空间相互作用是热点议题，然而关于其影响因素的非线性机制尚未被充分揭示。论文利用2018年腾讯迁徙大数据，以梯度提升决策树模型改进重力模型，分析距离衰减系数的非线性特征。研究发现：(1)城际联系的距离衰减系数呈非线性，偏依赖图曲线存在2个“平台区”(距离衰减系数接近于0)和2个“衰减区”(距离衰减系数约为2.8)，将直线距离替换为路网距离，2018年数据替换为其他年份数据后，结果仍然稳健；(2)与基准模型相比，航空、铁路及公路交通下的距离衰减系数非线性特征存在显著差异，春运期间的距离衰减系数基本保持一致。建议根据城际联系距离衰减规律布局都市圈和城市群，优化国土空间发展格局。

Nonlinear characteristics of distance-decay parameter in intercity links: Based on human mobility big data

Spatial interaction of human mobility is a heated topic for scholars, but little is known about the nonlinear relationship of the influencing factors in the spatial interaction. Using Tencent migration data in 2018 and employing the gradient boosting decision tree (GBDT) model, this study analyzed the nonlinear characteristics of the distance-decay parameter. The results show that: 1) In the annual data benchmark model the distance-decay parameter presents nonlinear characteristics; the partial dependence curve presents two "platform areas" where the distance-decay parameter is close to 0, and two "attenuation areas" where the distance-decay parameter is about 2.8; and the results are still robust using road-based distance or data from other years. 2) There are significant differences between the distance-decay parameters of aviation, railway, and road transportation. The distance-decay pattern during the Spring Festival is basically consistent with the annual benchmark model. We recommend to plan metropolitan areas and urban agglomerations according to the nonlinear characteristics of intercity connection influenced by distances-decay pattern, and optimize the layout of territorial development accordingly.