ＡＲＩＭＡ 模型在婴儿死亡率预测中的应用

摘要：目的　探讨应用自回归滑动平均混合模型（ａｕｔｏｒｅｇｒｅｓｓｉｖｅｉｎｔｅｇｒａｔｅｄ ｍｏｖｉｎｇａｖｅｒａｇｅ，ＡＲＩＭＡ） 预
测婴儿死亡率的可行性。方法　运用ＳＰＳＳ１６．０ 对１９９１－２０１２ 年山西省妇幼卫生年报婴儿死亡率建立
ＡＲＩＭＡ 模型，用所建模型比较预测值与实际值差异，并预测２０１３－２０１５年山西省婴儿死亡率。结果　模
型ＡＲＩＭＡ （１，２，０）较好地拟合了既往时间段的婴儿死亡率的时间序列， 模型自回归参数ＡＲ１＝
－０．７５４，犘＜０．０１，有统计学意义， 赤池信息准则（ＡＩＣ） ＝６８．２１３， 许瓦兹贝叶斯准则（ＳＢＣ） ＝
７０．２０４，模型残差为白噪声（犘＞０．０５）， 模型数学函数式为＾ 犢狋＝０．０６７＋１．２４６犢狋－１ ＋０．５０８犢狋－２ －
０．７５４犢狋－３，利用模型预测２０１３－２０１５ 年婴儿死亡率分别为４．７７‰、４．３２‰、３．９６‰。结论　ＡＲＩＭＡ 模
型能够较好地拟合婴儿死亡率的时间变化趋势，并用于短期预测未来婴儿死亡率。
关键词：婴儿死亡率；时间序列；ＡＲＩＭＡ 模型
中图分类号：Ｒ１９５．１　　文献标识码：Ａ　　文章编号：１００９ ６６３９ （２０１４）０３ ０２５６ ０４

Application of ARIMA model in the prediction of infant mortality rate

Objective To assess the feasibility of autoregressive integrated moving average（ARIMA）model in predicting the infant mortality rate（IMR）. Methods ARIMA model was established based on data of infant mortality rate analyzed by SPSS16.0and collected from Women and Children Health Statistic Report database during 1991to 2012in Shanxi.The predicted and actual IMR was compared and the model was used to predict IMR in 2013to 2015. Results The model of ARIMA（1,2,0）could better fit the time series of IMR in the study period.The autoregressive coefficient was statistically significant（AR1=-0.754,P〈0.01）（AIC =68.213,SBC=70.204）and the residuals was a white noise sequence（P〉0.05）.The mathematic function formula was^Yt=0.067＋1.246Yt-1＋0.508Yt-2-0.754Yt-3.The predicted IMRs for 2013,2014and 2015 were 4.77‰,4.32‰ and 3.96‰. Conclusions The ARIMA model can better fit the annual dynamic change of IMR,and may be used to predict IMR in the short-term.