不同数学模型对手足口病发病的预测效果分析与比较  

Application and Comparison of Different Mathematical Models in Prediction of Incidence Trend of Hand,Foot and Mouth Disease

作　　者：曹飞[1] 况荣华[2] 李辉[3] 刘明斌[4] 戚京城[3] 熊昌辉[4] 陈宝[1] 李譞超[1,5] 黄鹏[6] 

机构地区：[1]南昌大学研究生院医学部 [2]南昌大学期刊社 [3]南昌市疾病预防控制中心传染病防制科 [4]南昌市疾病预防控制中心科教信息科 [5]南昌330006 [6]南昌大学公共卫生学院流行病学教研室

出　　处：《南昌大学学报（医学版）》2014年第7期12-16,39,共6页Journal of Nanchang University：Medical Sciences

摘　　要：目的探讨不同数学模型对手足口病发病的预测效果并进行比较,为手足口病的监测和预防提供依据。方法收集南昌市2008—2012年手足口病疫情报告数据(以月为单位),分别应用曲线回归模型、单纯求和自回归滑动平均(ARIMA)模型、求和自回归滑动平均模型与多层感知神经网络(ARIMA-MLP)组合模型模拟手足口病的疫情变动轨迹,比较各模型的拟合效果,确定最优预测模型。结果南昌市2008—2012年手足口病发病率逐年上升,并呈现明显的季节趋势。对其发病趋势,3种模型均具有一定的预测能力,以ARIMA-MLP组合模型对手足口病月发病率的拟合效果最好(R2=0.908,MAE=3.06)。结论 ARIMA-MLP组合模型能够较好地拟合手足口病的发病趋势,对手足口病监测和预防具有一定意义。Objective To compare the efficacies of different mathematical models in the prediction of incidence trend of hand,foot and mouth disease,and to provide a scientific basis for the surveillance and prevention of hand,foot and mouth disease.Methods Epidemic data of hand,foot and mouth disease（2008-2012,Nanchang）were collected.The epidemic trajectory was simulated using curve regression model,autoregressive integrated moving average model（ARIMA）and ARIMA-multilayer perception neural network model（ARIMA-MLP）.The imitative effects the three models were compared to confirm optimal prediction model.Results The incidence of hand,foot and mouth disease increased year by year and presented obvious seasonal trend between 2008and 2012in Nanchang.All the three models had certain abilities to predict the incidence trend of hand,foot and mouth disease,especially ARIMA-MLP（R2=0.908,MAE=3.06）.Conclusion ARIMA-MLP model can simulate the epidemic trajectory of hand,foot and mouth disease,and has a certain significance for the surveillance and prevention of hand,foot and mouth disease.

关 键 词：手足口病 曲线回归模型 单纯求和自回归滑动平均模型 多层感知神经网络模型 

分 类 号：R181.2[医药卫生—流行病学]