摘要
目的分析北京市5岁以下儿童死亡率(U5MR)的长期变化趋势,探讨求和自回归滑动平均模型(ARIMA)预测U5MR的效果,并对2016-2020年北京市U5MR进行预测。方法基于北京市5岁以下儿童死亡监测网收集1992-2015年北京市5岁以下儿童死亡数据。以1992-2013年北京市U5MR为训练样本,拟合ARIMA模型。分别以1992-2013年和2014-2015年北京市U5MR为校验样本,对模型进行内外部验证。通过EViews 8.0和SPSS 19.0软件实现。结果 1992-2013年北京市U5MR呈下降趋势,22年间下降了81.50%。而2014和2015年略有回升,2年间上升了4.86%。构建最佳的死亡率预测模型为ARIMA(1,1,1):x_t=-0.445+1.509x_(t-1)-0.509x_(t-2)+α_t+0.999α_(t-1),R^2=0.982,MAPE=4.76%,AIC=2.15,BIC=2.30,满足白噪声。经内外部验证,实测值和预测值差异均无统计学意义(P_(内部)=0.401,P_(外部)=0.655)。利用此模型预测2016-2020年北京市U5MR,预测值分别为2.88‰、2.87‰、2.90‰、2.97‰和3.09‰。结论 1992-2013年北京市U5MR呈下降趋势,2014和2015年则略有回升。利用ARIMA(1,1,1)模型对北京市U5MR进行预测,效果较好。预测2016-2020年北京市U5MR呈现小幅上升趋势,提示应加大对北京市妇幼保健工作的投入和关注。
Objective To analyze the long-term change trend of under-5 mortality rate （U5MR） in Beijing, explore the effect of autoregressive integrated moving average （ARIMA） model in prediction of U5MR, and predict U5MR in Beijing from 2016 to 2020. Methods U5MR data in Beijing from 1992 to 2015 was collected based on U5MR Surveillance Network. U5MR from 1992 to 2013 was selected as training samples to fit ARIMA model. U5MR during 1992-2013 and 2014-2015 were selected as internal and external check-up samples to conduct internal and external validation, respectively. The data analysis and model establishment were accomplished through EViews 8. 0 software and SPSS 19. 0 software. Results U5MR in Beijing from 1992 to 2013 showed an overall decreasing trend by 81.50%. Whereas, U5MR in Beijing in 2014 and 2015 increased slightly by 4. 86%. ARIMA （ 1, 1, 1 ） model was the fittest predictive model, which was xt = -0. 445 ＋ 1. 509xt-1-0. 509xt-2 ＋αt ＋0. 999αt-1 , R^2 = 0. 982, MAPE = 4. 76% , AIC = 2. 15, BIC = 2. 30, the model satisfied the white noise （P〉0. 05 ） . There was no statistically significant difference between predictive value and actual value after internal and external validation （Piternal= 0. 401, P l = 0. 655 ） . The prediction values of U5MR in Beijing from 2016 to 2020 based on the model are 2. 88‰, 2. 87‰, 2. 90‰, 2. 97‰. and 3.09‰, respectively. Conclusion U5MR in Beijing from 1992 to 2013 showed an overall decreasing trend, whereas, U5MR in 2014 and 2015 increased slightly. ARIMA （ 1, 1, 1 ） model is appropriate to predict U5MR. U5MR may appear a slightly upward trend from 2016 to 2020 based on this model, which indicates that the investments and attention to maternal and child health care work should be enhanced in Beijing.
出处
《中国妇幼保健》
CAS
2017年第19期4605-4608,共4页
Maternal and Child Health Care of China
基金
国家重点研发项目(2016YFC0900603)
