基于SARIMA-SVR组合模型的 空气质量指数预测

空气质量指数(AQI)在波动中既具有整体的时间序列线性特征和明显的季节性波动周期,又具有多种因素影响的不确定性,为了提高AQI的预测精度,基于Ri386 3.3.3和Matlab R2014a两种编程软件,提出了一种同时具有线性和非线性的复合特征的时间序列预测模型——SARIMA-SVR组合模型。以太原市2014年1月—2019年7月的AQI月均值数据为基础,利用SARIMA时间序列模型进行线性预测,利用SVR模型对残差进行非线性预测,加和得到组合预测模型的预测结果,分析比较SARIMA,SVR和SARIMA-SVR这3种模型的预测结果和平均绝对百分比误差。结果表明,组合预测模型发挥了2种模型各自的优势,相较于单一预测模型的预测结果而言,其预测精度更高,稳定性更好。通过此模型得到的空气质量预测结果不仅可为人们的日常生活提供指导,而且可为大气污染的防治工作提供科学依据和借鉴意义。

Air quality index prediction based on SARIMA-SVR combined model

Air quality index (AQI) both has volatility of time series of the whole linear features, obvious seasonal fluctuation cycle, at the same time has a variety of factors of uncertainty. In order to improve the prediction accuracy of AQI, based on Ri386 3.3.3 and Matlab R2014a programming software, this paper proposes a composite characteristics of both linear and nonlinear time series prediction model,namely SARIMA-SVR combined model. Based on the monthly average data of AQI from January 2014 to July 2019 in Taiyuan, SARIMA time series model is first used for linear prediction, then SVR model is used for non-linear prediction of residual, and finally the combined prediction model is added and obtained. By analyzing and comparing the prediction results and average absolute percentage errors of SARIMA, SVR and SARIMA-SVR models, the results show that the combined prediction model gives full play to the advantages of the two models, and its prediction accuracy is higher and its stability is better than that of the single prediction model. The prediction results of air quality by this model can provide reference for the prevention and control of air pollution.