摘要
人口空间化是实现人口统计数据与其他环境资源空间数据融合分析的有效途径。选取地形地貌、地表覆盖、植被覆盖、河流水系、交通可达、经济活跃、生活便利作为影响成都市人口分布的变量,首先根据主成分分析将变量降维,然后利用机器学习对成都市2020年人口进行空间化,将模拟结果与WorldPop和人口普查数据进行对比,得出基于随机森林模型的模拟结果准确度最高,最后通过随机森林的因子重要性分析成都市人口空间分布的影响因素。结果表明:研究结果在街道尺度整体精度均达到90%,均优于WorldPop;POI是高精度人口分布的最重要指标,夜间灯光、坡度和路网密度等对成都市人口分布也有重要作用。POI可有效提升人口空间化结果的准确性,主成分分析是综合POI因子的可行方案,随机森林模型能够有效融合多源信息,为城市层面的精细人口空间化研究提供参考。
Population spatialization is an effective method for integrating and analyzing population statistics data with other spatial environmental resources.In the case of Chengdu City,variables such as landform,land cover,vegetation cover,river system,accessible transportation,economic activity,convenient life were identified as factors influencing population distribution.Initially,the dimensionality of these variables was reduced using principal component analysis.Subsequently,machine learning techniques were employed to spatialize the 2020 population of Chengdu City.The simulation results were then compared with data from WorldPop and census records.It was determined that the random forest model produced the most accurate simulation results.Furthermore,an analysis of the factors affecting the spatial distribution of Chengdu’s population was conducted using random forest’s factor importance measure.The findings indicated that at a street scale level,the overall accuracy of this study reached 90%,surpassing that of WorldPop.Additionally,it was revealed that POI is a crucial determinant in achieving high-precision population distribution;while night light intensity,slope gradient and road network density also significantly impact population distribution in Chengdu City.POI effectively enhances population spatialization outcomes;principal component analysis offers a viable approach for synthesizing POI-related factors;and lastly,random forest model successfully integrates multi-source information providing valuable insights for fine-scale urban population spatialization studies.
作者
张晓荣
周垠
张娜
于儒海
蒋源
ZHANG Xiaorong;ZHOU Yin;ZHANG Na;YU Ruhai;JIANG Yuan(Chengdu Institute of Planning&Design,Chengdu 610000,China)
出处
《地域研究与开发》
CSSCI
北大核心
2024年第5期173-180,共8页
Areal Research and Development
关键词
人口空间化
机器学习
人口分布
数据降维
随机森林
多源数据融合
成都市
population spatialization
machine learning
population distribution
data dimensionality reduction
random forest
multi-source data fusion
Chengdu City
