空气质量数据的校准研究

Calibration of Air Quality Data

针对2019年全国大学生数学建模竞赛D题空气质量数据校准问题,在一些简化的假设下,本研究利用SPSS软件对经过预处理的“两尘四气”的国控点数据和自建点数据进行探索性数据分析,包括描述性数据分析、配对样本T检测,得出两组数据之间存在着一定的差异性;通过回归性分析,得出PM2.5与PM10互为最大影响因素,对CO的值影响最大的因素是O_(3)和风速等结论;最后通过建立多元线性回归模型对自建点数据进行校准,选用国控点数据与自建点数据中的2/3数据,建立多元线性回归模型,选取后1/3数据作为样本来检验,计算出检验结果并分析得出与国控数据的Pearson值均大于0.5,证明该模型误差很小,具有一定可靠性,将自建点数据分别带到所建模型中,计算得出校准数据。

In view of the 2019 national college students’ mathematical modeling contest D air quality data calibration problem, under some simplified assumptions, this paper using SPSS software to after pretreatment of four gas“dust”two countries control point data and self-built point data of exploratory data analysis, including descriptive data analysis, paired sample T test, it is concluded that there is certain difference between two groups of data. Through correlation coefficient analysis and regression analysis, it is concluded that PM2.5 and PM10 are the biggest influencing factors for each other, while O_(3)and wind speed are the factors that have the biggest influence on CO value. Finally through multiple linear regression model is established to calibration of self-built point data, choose the control point data and self-built two-thirds of data points in the data, multiple linear regression model is set up, after selecting a third data as sample to test, analysis and test results by the state-controlled data Pearson value were greater than 0.5, proved that the model error is very small, has a certain reliability, will build some data into the model, respectively, calculated from the calibration data.