自回归移动平均模型在预测PM2.5中的应用  

Forecast of PM2.5 with the ARMA model

作　　者：刘宗伟[1] 周彩丽[2] 马冬梅 邱玉刚[1] 翟庆峰[1] 刘春兰 

机构地区：[1]潍坊医学院公共卫生与管理学院,山东潍坊261053 [2]潍坊医学院临床医学院,山东潍坊261053 [3]战略支援部队兴城疗养院接诊科,辽宁葫芦岛125105 [4]潍坊市奎文区卫生计生监察大队,山东潍坊261000

出　　处：《预防医学论坛》2016年第8期582-584,共3页Preventive Medicine Tribune

摘　　要：目的应用自回归移动平均（Auto-regressive Moving Average Model）模型,建立潍坊市PM2.5的日均浓度预测模型。方法利用潍坊市2013年12月2日至2016年9月9日的历史PM2.5日均浓度数据,采用条件最小二乘法确定模型参数,模型阶数确定后,建立PM2.5日均浓度预测模型。结果对模型的各个参数进行检验发现,各参数估计值的P值均〈0.05;对建立的模型进行残差的白噪声检验,χ~2检验统计量的P值均〉0.05,据此建立ARMA（1,3）模型,模型表达式为：χ_t=138.188 8＋（1-0.394 23β-0.367 09β~2-0.146 84β~3）ε_t/（1-0.999 6β）,并预测了PM2.5的未来日均浓度。结论 ARMA（1,3）模型可用于预测潍坊市PM2.5的日均浓度变化趋势。Objective To explore the application of auto-regressive moving average（ARMA）model of time series in forecasting the daily average concentration of PM2.5. Methods ARMA model was fitted with data of daily reported cases in Weifang from December 2in 2013 to September 9in 2016 for PM2.5.Parameters were estimated according to conditional least square method.ARMA model was established for daily average concentration of PM2.5in Weifang after index was identified. Results After testing for each parameters of model,the estimated value of parameters was statistically significant.P-values fort-test of all coefficients were all below 0.05.Then the white noise check was used to test the residuals of model,the results showed that the P-values of statistics were all greater than 0.05 by chi-square test.ARMA was identified to fit and forecast daily average concentration of PM2.5.Model equations wers χ_t=138.188 8＋（1-0.394 23β-0.367 09β~2-0.146 84β~3）ε_t/（1-0.999 6β）. Conclusion The model of ARMA can be used to forecast the changes of daily average concentration of PM2.5in Weifang city.

关 键 词：时间序列 自回归移动平均模型 PM2.5 预测 

分 类 号：R195.1[医药卫生—卫生统计学]