地理科学 ›› 2021, Vol. 41 ›› Issue (10): 1763-1772.doi: 10.13249/j.cnki.sgs.2021.10.008

• • 上一篇    下一篇

基于随机森林模型的中国流动人口社会融合空间差异及影响因素

郑茹敏1(), 梅林1,2,*(), 姜洪强1, 付占辉3, 甄硕1   

  1. 1.东北师范大学地理科学学院,吉林 长春 130024
    2.长春财经学院管理学院,吉林 长春 130111
    3.河南大学地理与环境学院,河南 开封 475004
  • 收稿日期:2020-10-12 修回日期:2021-01-22 出版日期:2021-10-25 发布日期:2021-12-08
  • 通讯作者: 梅林 E-mail:ruminzheng@163.com;meil682@nenu.edu.cn
  • 作者简介:郑茹敏(1992−),女,山西长治人,博士研究生,主要从事人口地理与旅游地理研究。E-mail: ruminzheng@163.com
  • 基金资助:
    国家自然科学基金项目(41971202)

Spatial Differences and Impact Factors of Migrant Integration in China Based on Random Forest Model

Zheng Rumin1(), Mei Lin1,2,*(), Jiang Hongqiang1, Fu Zhanhui3, Zhen Shuo1   

  1. 1. School of Geographical Sciences, Northeast Normal University, Changchun 130024, Jilin, China
    2. School of Management, Changchun University of Finance and Economics, Changchun 130111, Jilin, China
    3. College of Geography and Environmental Science, Henan University, Kaifeng, 475004, Henan, China
  • Received:2020-10-12 Revised:2021-01-22 Online:2021-10-25 Published:2021-12-08
  • Contact: Mei Lin E-mail:ruminzheng@163.com;meil682@nenu.edu.cn
  • Supported by:
    National Natural Science Foundation of China(41971202)

摘要:

基于2017年流动人口动态监测数据,采用探索性空间数据分析和随机森林回归方法,分析中国289个地级及以上城市流动人口社会融合的空间分异及影响因素。结果表明:① 中国城市流动人口社会融合整体水平较低,空间分布地带性特征不明显;特大、超大城市流动人口社会融合分别以较低、中等水平为主,省会城市、直辖市流动人口社会融合分别以较低、中等水平为主,其余城市多数为中等水平;城市社会融合空间分布格局表现为以峰值为中心向外圈层递减和以低值为中心向外圈层递增的特征。② 随机森林模型的拟合度和回归精度比多元线性回归更高,能够更好地解释流动人口社会融合的非线性影响机制。③ 个人教育水平、家属是否随同、流入城市经济结构、流动人口的户口类型、流入城市公共设施以及流动人口在流入地居留时间依次为六大重要影响因子,它们对流动人口社会融合的作用呈现复杂性和非线性。④ 需要根据影响因子在不同阈值内的作用方式,有针对性地对流动人口及流入城市进行管理和调控。

关键词: 流动人口, 社会融合, 随机森林模型

Abstract:

Promoting the migrant integration is an effective way to realize people-oriented urbanization. Using the data of 2017 China Migrants Dynamic Survey, the index system of migrant integration is constructed. The spatial differences and influencing factors of migrant integration of 289 cities in China are analyzed by using the methods of exploratory spatial data analysis and Random Forest Model. The results show that: 1) The overall level of migrant integration is low, and the zonal spatial distribution characteristics are not obvious. The level of migrant integration in mega- and super-cities is mainly at a low and medium level, and the that in provincial capitals and municipalities is mainly at a low and medium level respectively, and most of the remaining cities are at a medium level. The spatial distribution pattern is characterized by decreasing circle around the peak center and increasing circle around the low center. 2) The random forest model has higher fitting degree and regression precision than the multiple linear regression, and it is more suitable for the analysis of the non-linear mechanism of migrant integration. 3) According to the importance of factors influencing migrant integration, the top six important influencing factors are: migrants’ individual education level, whether the family members migrate with migrants, the economic structure of the receiving city, the type of household registration of migrants, the public facilities of the receiving city and migrants’ residence time in the receiving city. These six factors have complex and nonlinear effects on migrant integration. 4) It is necessary to take targeted strategies from migrants and receiving cities according to the importance and effects of influence factors.

Key words: domestic migrants, migrant integration, Random Forest Model

中图分类号: 

  • K901