摘要
【目的】传统的空气质量计算方法已经难以满足当前社会的需求,基于机器学习的空气质量模型的分析与研究有助于提高空气质量预测的准确性和及时性。【方法】选取郑州市2019年全年空气质量和气象的逐小时数据,分别分析了空气质量与空气污染物因子的相关性以及空气污染物因子与气象要素的相关性。以空气污染物和气象要素作为输入因子,采用机器学习的方法建立多个回归模型,最后通过对回归模型的评估,根据评估的结果选取合适的预测模型。【结果】通过对回归模型评估指标的三种方法进行对比发现,随机森林模型效果较好,验证了机器学习的方法非常适合于空气质量的预测。【结论】本研究适应于站点的模型建立与预测,下一步应继续进行深度研究,从站点来繁衍出格点的实况预测模型。
[Purposes]Traditional air quality calculation methods have been unable to meet the needs of the current society.The analysis and research of air quality model based on machine learning is helpful to improve the accuracy and timeliness of air quality prediction.[Methods]Hourly data of air quality and meteorology of Zhengzhou in 2019 were selected to analyze the correlation between air quality and air pollutant factors as well as the correlation between air pollutant factors and meteorological factors.With air pollutants and meteorological elements as input factors,multiple regression models are established by machine learning method.Finally,through the evaluation of regression models,an appropriate pre⁃diction model is selected according to the evaluation results.[Findings]Through the comparison of three methods of regression model evaluation index,random forest model is better.Thus,it is verified that ma⁃chine learning method is very suitable for air quality prediction.[Conclusions]This study is suitable for model establishment and prediction of the site.In the next step,in-depth research should be carried out to reproduce the real prediction model of unusual points from the site.
作者
朱茜
ZHU Qian(Meteorological Data Center of Henan province,Zhengzhou 450003,China)
出处
《河南科技》
2023年第3期95-99,共5页
Henan Science and Technology
基金
河南省农业气象保障与应用技术重点实验室应用技术研究基金项目“基于气象大数据云平台的环境气象数算一体应用研究”(KM202116)。
关键词
空气质量
机器学习
回归模型
air quality
machine learning
the regression model
