测绘通报 ›› 2020, Vol. 0 ›› Issue (1): 76-81.doi: 10.13474/j.cnki.11-2246.2020.0016

• 学术研究 • 上一篇    下一篇

集成地理探测器与随机森林模型的城市人口分布格网模拟

成方龙1, 赵冠伟1,2, 杨木壮1,2, 刘月亮1, 李芳1   

  1. 1. 广州大学地理科学学院, 广东 广州 510006;
    2. 广州大学国土资源与海岸带研究所, 广东 广州 510006
  • 收稿日期:2019-04-08 修回日期:2019-10-19 发布日期:2020-02-10
  • 通讯作者: 赵冠伟。E-mail:zgw11124@163.com E-mail:zgw11124@163.com
  • 作者简介:成方龙(1992-),男,硕士生,主要研究方向为地理时空数据分析。E-mail:784628444@qq.com
  • 基金资助:
    国家自然科学基金(41671175;41671430);广东省自然科学基金(2017A030313240);广东省高等学校优秀青年教师培养项目(YQ2015127)

Simulation of urban population distribution grid by integrating geodetector and random forest model

CHENG Fanglong1, ZHAO Guanwei1,2, YANG Muzhuang1,2, LIU Yueliang1, LI Fang1   

  1. 1. School of Geographical Science, Guangzhou University, Guangzhou 510006, China;
    2. Institute of Land Resources and Coastal Zone, Guangzhou University, Guangzhou 510006, China
  • Received:2019-04-08 Revised:2019-10-19 Published:2020-02-10

摘要: 精细尺度的城市人口分布信息是城市资源配置和综合管理的重要依据。本文以广州市越秀区、荔湾区、天河区、海珠区、白云区及黄埔区作为研究区域,基于人口统计、夜间灯光、兴趣点及土地利用等多源数据,利用地理探测器识别人口分布的影响因子,运用随机森林模型开展人口分布空间格网模拟研究。研究结果表明,与传统的相关分析相比,地理探测器能够更为准确地识别人口空间分布的重要影响因子。基于随机森林模型的人口分布格网模拟结果与街道(镇)实际人口的相关系数为0.774,平均相对误差约为30%。相比基于线性回归模型的模拟结果,随机森林模型的精度有明显提高。

关键词: 人口分布, 格网, 模拟, 随机森林, 地理探测器

Abstract: Fine population distribution is important to urban resource allocation and management.In this paper,we take the Yuexiu District, Liwan District, Tianhe District, Haizhu District, Baiyun District and Huangpu District of Guangzhou city as the research areas,and base on multi-source data such as demography, night lighting, interest points and land use,using the geodetector to identify the influencing factors of population distribution,and simulate the population distribution grid by using random forest model.The results show that compared with the traditional correlation analysis, the geodetector can identify the important factors of spatial distribution of population more accurately.The correlation coefficient between the results of population distribution grid simulation based on random forest model and the actual population of streets (towns) is 0.774, with an average relative error of about 30%.Compared with the simulation results based on linear regression model, the accuracy of the stochastic forest model is significantly improved.

Key words: population distribution, grid, simulation, random forest, geodetector

中图分类号: