摘要
为验证多元统计方法在PM2.5分析及预测方面的适用性,以合肥地区为例,收集了2015全年的PM2.5数据,借助于统计分析软件R进行了相关实验。通过PM2.5与各个影响因素之间的散点图,发现部分影响因素和PM2.5存在着较强的线性关系,据此建立关于PM2.5的多元线性回归模型。在保证各个变量不相关、独立条件下,对模型进行了验证。根据验证结果,选用了逐步回归分析方法得到了一个新模型。根据调整R2最大的原则确定了最终模型。最后选用了RMSE、MAE和Theil不相等系数对模型的预测效果进行了检验,模型整体预测效果较好。
In order to verify the applicability of multivariate statistical methods in PM2.5 analysis and prediction, taking Hefei area as an example, the PM2.5 data of 2015 were collected, and related experiments were carried out by means of statistical analysis software (R). Firstly, through the scatter diagram between PM2.5 and each influencing factor, it is found that there is a strong linear relationship between some factors and PM2. 5, and next the multiple linear regression model of PM2.5 is established. Under the condition that the variables are not relevant and independent, the prediction model is validated. Based on the result of verification, a new model is obtained by stepwise regression analysis. Then, the final model is determined according to the principle of adjusting the maximum R2. Finally, the RMSE, MAE and Theil unequal coefficients are used to test the prediction effect of the model, and the overall prediction results of the model are better.
出处
《佳木斯大学学报(自然科学版)》
CAS
2018年第1期96-99,118,共5页
Journal of Jiamusi University:Natural Science Edition
基金
安徽省高校自然科学重点项目(KJ2015A309)
安徽省大学生创新训练项目(201612216082)
国家大学生创新训练项目(201712216019)
关键词
PM2
5
R软件
多元线性回归模型
预测
PM2.5
R software
multiple linear regression model
prediction