基于集成学习与地理空间大数据的中国第七次人口普查100米网格化人口数据集

2020年中国第七次人口普查100米网格化人口数据集通过堆叠集成学习与地理空间大数据技术生成。研究首先收集了县级和乡级普查数据及十项100米分辨率相关协变量作为输入数据集，选用随机森林、XGBoost和LightGBM三种主流机器学习算法作为基模型，构建并训练堆叠集成学习框架，最终生成了中国高精度网格化人口数据集。经乡镇级普查测试数据验证，该数据集（R²=0.8936）的精度显著优于现有的WorldPop（R²=0.7427）和LandScan（R²=0.7165）产品。本数据集关联发表于《地球系统科学数据》期刊的论文《陈跃洪、徐聪聪、葛勇、张晓翔、周亚男. 基于集成学习与大地理空间数据的中国七普100米网格化人口数据集, 2024》。

The 100-meter gridded population dataset derived from China's 7th National Population Census in 2020 was generated via stacked ensemble learning and geospatial big data technologies. This study first collected county-level and township-level census data along with ten 100-meter resolution related covariates as the input dataset, then adopted three mainstream machine learning algorithms—Random Forest, XGBoost, and LightGBM—as base models to construct and train a stacked ensemble learning framework, ultimately producing China's high-precision gridded population dataset. Validated against township-level census test data, this dataset achieved an R² score of 0.8936, whose accuracy significantly outperforms existing WorldPop (R²=0.7427) and LandScan (R²=0.7165) products. This dataset is associated with the paper published in *Earth System Science Data*: "Chen Yuehong, Xu Congcong, Ge Yong, Zhang Xiaoxiang, Zhou Yanan. China's 100-meter Gridded Population Dataset from the 7th National Population Census Based on Ensemble Learning and Large Geospatial Data, 2024".