应用3种回归模型预测手足口病周发病数

目的 探索分析手足口病周数据的统计学方法，提升手足口病预测能力。方法 中国疾病预防控制信息系统导出2008年第1周至2014年第14周北京市通州区手足口病周发病数。采用SPSS 17.0 软件进行自回归、季节性自回归与混合Serfling 回归模型拟合。结果 自回归、季节性自回归、混合Serfling回归3种模型对2008年第1周至2014年第14周实际发病数进行拟合，回归方程R2分别是0.907、0.917、0.919，所得残差经Ljung-Box检验均是白噪声；以所得回归方程对2014年第15周至第38周实际发病数进行预测，3种模型的平均绝对百分比误差（MAPE）分别为:18.67%、18.43%、17.12%。 结论 混合Serfling回归模型预测效果最优。

The prediction of the weekly incidence of hand-foot-mouth disease (HFMD) by three regression models

Objective The aim of this study was to search the model for the prediction of weekly incidence of Hand-Foot-Mouth Disease (HFMD), and to strengthen the ability of the prediction of HFMD. Methods SPSS 17.0 software was used to run the auto-regression model, the seasonal auto-regression model, and the mixed Serfling model on the HFMD incidence data from the 1st week of 2008 to the 14th week of 2014 in China Information System for Disease Control and Prevention. Results The R2 of the auto-regression model, the seasonal auto-regression model, and the mixed Serfling model to fit the incidence from the 1st week of 2008 to the 14th week of 2014 were 0.907, 0.917, and 0.919, respectively, and their residuals were all white noises under the Ljung-Box test. The mean absolute percentage errors (MAPE) of the auto-regression model, the seasonal auto-regression model, and the mixed Serfling model to predict the incidence from the 15th week of 2014 to the 38th week of 2014 were 18.67%, 18.43%, and 17.12%, respectively. Conclusion The mixed Serfling model had the best effect in the prediction.