摘要
融合多源数据可以得到精细尺度城市格网人口数据,但输入数据的采样偏差、空间覆盖和时效性等因素容易造成人口估算误差。针对上述问题,以郑州市市辖区为研究区,设计一种基于无线网络数据的城市人口空间化方法。通过整合兴趣点(POI)数据和借助广告科技生态获取的无线网络数据构建人口吸引力指标模型,再基于随机森林和地理加权回归两种模型进行人口空间化实验,生成100 m分辨率的城市格网人口数据。结果表明:基于该方法的格网人口数据精度均有显著提升,验证了人口吸引力指标的有效性。研究区呈现“核心—边缘”以及圈层递减的典型城市人口分布结构。在乡镇/街道层面精度评价中,各类精度误差在数值上均优于公开数据集WorldPop和LandScan。
Fine-scale urban grid population data is obtained by fusing data from multiple sources,yet factors like sampling bias,spatial coverage,and temporality of input data often lead to population esti-mation errors.To address these issues,an urban population mapping method based on WiFi data is de-signed,with Zhengzhou city area as the study region.A population attraction indicator model is con-structed by integrating point of interest(POI)data with WiFi data sourced from ad tech ecosystems.Population mapping experiments are then conducted using two models:random forest and geographi-cally weighted regression,to produce 100 m resolution urban grid population data.The results indicate significant improvement in the accuracy of grid population data generated by using the proposed meth-odology,validating the effectiveness of the population attractiveness indicator.The study area exhibits a typical“core-edge”urban population distribution structure with decreasing circles.At the township/street level accuracy assessment,the method demonstrates numerically lower accuracy errors com-pared to publicly available datasets such as WorldPop and LandScan.
作者
李帅
徐青
朱新铭
李雪嫚
闫旭强
LI Shuai;XU Qing;ZHU Xinming;LI Xueman;YAN Xuqiang(Information Engineering University,Zhengzhou 450001,China)
出处
《信息工程大学学报》
2024年第6期717-724,共8页
Journal of Information Engineering University
关键词
无线网络数据
兴趣点
人口空间化
人口吸引力指标
人口分布格局
WiFi data
point of interest
population mapping
population attraction indicators
popu-lation distribution pattern
