摘要
基于2014—2021年京津冀地区14个城市的污染物数据(PM2.5、CO)进行插补方法的性能评估实验,发现相较于常规插补方法,概率函数聚类插补(Probabilistic Functional Clustering Imputation,PFCI)方法的性能更优。在此基础上,利用PFCI方法对2021年北京市35个监测站点的6种污染物缺失数据进行实证应用。结果表明:针对大规模稀疏空气质量数据,PFCI方法能够较好识别其潜在变化模式,提高缺失值插补的精度。
This paper carried out the performance evaluation experiment of imputation methods base on the pollutant data(PM2.5,CO)of 14 cities in Beijing-Tianjin-Hebei region from 2014 to 2021,and it was found that the Probabilistic Functional Clustering Imputation(PFCI)method had better performance than the conventional imputation methods.On this basis,the PFCI method is used to empirically apply the missing data of 6 pollutants from 35 monitoring stations in Beijing in 2021.The results show that PFCI method can improve the accuracy of missing data imputation by accurately identifying its potential change pattern for large-scale sparse air quality data.
作者
高海燕
马文娟
李唯欣
张悦
Gao Haiyan;Ma Wenjuan;Li Weixin;Zhang Yue(School of Statistics,Lanzhou University of Finance and Economics,Lanzhou Gansu 730020,China)
出处
《河北环境工程学院学报》
CAS
2023年第5期73-82,共10页
Journal of Hebei University of Environmental Engineering
基金
国家社会科学基金项目(19XTJ002)
甘肃省自然科学基金项目(23JRRA1186)
甘肃省优秀研究生“创新之星”项目(2023CXZX-703,2022CXZX-701)。
关键词
稀疏空气质量数据
函数型数据分析
概率函数聚类插补
实证研究
sparse air quality data
functional data analysis
probabilistic functional clustering im-putation
empirical research
