SARIMA模型在流行性腮腺炎发病预测中的应用  

Application of SARIMA model in predicting the incidence of mumps

作　　者：李润滋 章涛[1] 梁玉民[2] 罗成[1] 蒋正[1] 薛付忠[1] 刘言训[1] 刘静[1] 李秀君[1] 

机构地区：[1]山东大学公共卫生学院生物统计学系,山东济南250012 [2]济宁市疾病预防控制中心,山东济宁272000

出　　处：《山东大学学报（医学版）》2016年第9期82-86,96,共6页Journal of Shandong University:Health Sciences

基　　金：山东省科技发展计划(2014GGH218019);病原微生物生物安全国家重点实验室开放课题(SKLPBS1453)

摘　　要：目的利用SARIMA模型预测未来山东省济宁市流行性腮腺炎发病情况,为流行性腮腺炎防控提供决策依据。方法收集山东省济宁市2009年1月至2013年7月流行性腮腺炎月发病数据资料,利用时间序列分析方法,构建SARIMA模型,并对2013年8月至12月的发病数资料进行预测。结果济宁市2009至2013年共报告流行性腮腺炎病例数8 520例,且发病具有明显的周期性和季节性特征。最终建立的最优模型为SARIMA(0,1,1)(0,1,1)_(12),赤池信息准则(AIC)为74.45,且通过了统计学检验,模型残差为白噪声。实际月发病数与拟合月发病数进行相关性分析结果显示为显著性相关(r=0.75,P<0.000 1)。对2013年8月至12月发病数进行预测,均在95%置信区间内,且与实际发病数变动的趋势一致,验证了模型合理性。结论 SARIMA模型能较好地拟合济宁市流行性腮腺炎月发病数动态变化,可用于流行性腮腺炎的短期预测。Objective To predict the incidence of mumps with autoregressive integrated moving average （SARIMA） model so as to provide scientific guidance for its prevention and control. Methods Time-series data of monthly mumps cases from Jan. 2009 to July 2013 were analyzed using SARIMA model and predictive model was established to predict the incidence from August to December 2013. Results From 2009 to 2013, a total of 8,520 cases of mumps were reported in Jining City. Eventually the optimal model of SARIMA （0,1,1） （0,1,1） 12 was established, and the information criterion （AIC） was 74. 45. Parameters estimated were statistically significant, and residuals were white noise sequence. Monthly mumps cases from January 2009 to July 2013 were used for model fitting and the monthly mumps cases from Aug. to Dec. 2013 predicted by the optimal model were within the 95% confidence interval, and were consistent with the trend of the actual incidence, which demonstrated the rationality of the model. Correlation between actual case number and fitted case number was statistically significant （ r = 0. 75, P 〈 0. 000 1 ）. Conclusion SARIMA model can fit the incidence of dynamic change of mumps, and can be used to make short-term prediction and to provide scien- tific evidence for the prevention and control of mumps.

关 键 词：时间序列分析 季节性差分自回归移动平均模型 流行性腮腺炎 预测 

分 类 号：R181.25[医药卫生—流行病学] R183.3[医药卫生—公共卫生与预防医学]