基于多源数据与XGBoost模型的上海市人口空间化研究

Spatialization Research on Shanghai's Population Based on Multi-source Data and XGBoost Model

以特大城市上海为研究对象,融合珞珈一号夜间灯光、兴趣点、DEM、分空间土地利用、河流等多源数据,建立多源特征数据库,基于GridSearchCV得分评价构建XGBoost模型,实现上海市100 m×100 m人口空间化,并与Worldpop人口数据集进行精度对比分析。结果表明,上海市人口分布呈多中心分布特征,LJ-01夜光数据与POI数据对人口空间化起重要的辅助作用,土地利用数据虽然对人口空间化重要性程度不高,但用地空间划分同样能体现出不同用地空间功能的差异,且本研究结果精度(R^(2)=0.98)高于Worldpop人口数据集(R^(2)=0.78),说明XGBoost模型具有较高的可靠性,可为其他大型城市人口空间化研究提供参考。

Taking the mega city Shanghai as the research object,we fuse multi-source data such as nighttime lighting of LJ-01,points of interest,DEM,sub-spatial land use,rivers,etc.,to establish a multi-source feature database,and based on the GridSearchCV score evaluation,the XGBoost model was constructed to realize the 100 m×100 m population spatialization in Shanghai,and the accuracy is compared with the Worldpop population dataset.The results show that the population distribution in Shanghai is characterized by multi-center distribution.LJ-01 nighttime lighting and POI data play an important auxiliary role in population spatialization.Although land use data is not of high importance to population spatialization,the spatial division of land use is equally important.It can reflect the differences in spatial functions of different land uses,and the accuracy of the results of this study(R^(2)=0.98)is higher than that of the Worldpop population dataset(R^(2)=0.78),indicating that the XGBoost model has high reliability and can provide reference for other large urban population spatialization research.