The current atmospheric PM2.5 concentrationprediction model parameters are not well fitted,which leads to large prediction errors.A predictionmodel of atmospheric PM2.5 concentration based ontime series analysis is proposed.We test the station-arity of the original atmospheric PM2.5 data series,obtain the sample autocorrelation function and sam-ple partial autocorrelation function of the observa-tion sequence,introduce the AIC criterion,and esti- mate the unknown parameters in the model.Basedon the model optimization,we predict the atmos- pheric PM2.5 concentration.Taking a city as the ob-ject,we collect the PM2.5 concentration data in theatmosphere of the city from January 1,2019 to De-cember 28,2019.By determining the data scale,pro-cessing abnormal values,completing missing data, and data standardization processing,we preher ePM2.5 concentration of the city from December 29, 2019 to December 31,2019.The experimental re-sults can be obtained: the original data series of at-mospheric PM2.5 is not stable,and the autocorrela-tion coefficient is always greater than 0.Using thisdata sequence for model testing,the residual distri- bution range is -2~+2,the residual autocorrelationrange is -0.15~+0.15,the p value is always within 1, and the above parameters are within the ideal range,indicating that the fitting effect is good.
展开▼
