ARIMA乘积季节模型在日喀则市结核病疫情预测中的应用  

Application of multiple seasonal ARIMA model in prediction of tuberculosis precalence in Shigatse

作　　者：尹亭亭 李志娟[2] 巴桑片多 张倍 次仁加布 旺杰 强巴桑珠 李林华 胡军 YIN Ting-ting;LI Zhi-juan;Basangpianduo;ZHANG Bei;Cirenjiabu;WANG Jie;Qiangbasangzhu;LI Lin-hua;HU Jun(School of Public Health and Management,Weifang Medical University,Weifang Shandong,261053,China;Party Committee Office,Shandong Blood Center,Jinan Shandong,250014,China;Central Office and Tuberculosis Control Department,Disease Control and Prevention Center of Shigalse,Shigause Tibet,857000,China;AIDS Prevention office,Disease Control and Prevention Center of Shandong,Jinan Shandong,250014,China)

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]山东省血液中心党委办公室,山东济南250014 [3]日喀则市疾病预防控制中心中心办公室及结核病防治科,西藏日喀则857000 [4]山东省疾病预防控制中心艾滋病防制所,山东济南250014

出　　处：《职业与健康》2020年第5期634-637,共4页Occupation and Health

基　　金：西藏自治区自然科学基金(XZ2017ZRG-83)。

摘　　要：目的通过构建的自回归滑动平均混合(ARIMA)乘积季节模型预测日喀则市结核病月发病情况,通过比较模型预测值与实际值来评估该模型的预测效果。方法根据日喀则市2010年1月-2016年12月的结核病月发病例数构建ARIMA乘积季节模型,利用该模型预测2017年1-12月的结核病月发病情况,通过比较预测值与实际值来评价拟合模型的预测效果。结果最优模型为ARIMA(1,1,1)(1,1,0)12,其参数均通过统计学检验(均P<0.05),残差序列为白噪声序列(P>0.05),其赤池信息准则(AIC)=679.48,许瓦兹贝叶斯准则(SBC)=686.27,拟合优度相对最好。2017年1-12月的预测值与实际值基本吻合,实际值均落在95%CI内,预测效果较好。结论ARIMA(1,1,1)(1,1,0)12可用于短期预测日喀则市结核病疫情,预测效果较好,建议要及时根据数据更新及其他的因素更新模型,以确保模型的预测价值。Objective To predict the monthly data of tuberculosis in Shigatse by establishing a multiple seasonal Autoregressive Integrated Moving Average(ARIMA)model,to evaluate its predicting ability by comparing the predicted value of the model and the actual value.Methods Based on the monthly data of the Shigatse from January 2010 to December 2016,the multiple seasonal ARIMA model was established,which was used in predicting the number of cases from January to December 2017,and to evaluate its forecasting ability by comparing the predicted and actual value of the model.Results The results indicated that predicting effect of the model ARIMA(1,1,1)(1,1,0)12 was prefect,whose parameters were statistically different(all P<0.05).The residual sequence was a white-noise sequence(P>0.05).The Akaike Information Criterion(AIC)was 679.48,the Schwarz Bayesian Criterion(SBC)was 686.27.Goodness of fit was the best.The predicting value for January 2017 to December 2017 were generally consistent with the actual values.Conclusion ARIMA(1,1,1)(1,1,0)12 can be used to predict the epidemic situation of tuberculosis in Shigatse in a short-term.The model has good popularization and application value.However,to update the model in time based on data updates and other existing factors was recommended to ensure the predictive value of the model.

关 键 词：ARIMA乘积季节模型 结核病 月发病例数 预测 

分 类 号：R181.8[医药卫生—流行病学]