河北省流行性腮腺炎周报告疫情时间序列比较分析

Comparative analysis of time series of weekly mumps reports in Hebei Province

目的对河北省流行性腮腺炎(简称“流腮”)周报告疫情使用不同的时间序列方法进行比较分析,选择合适的方法对其做出更便捷的预测。方法利用河北省2004年第1周至2020年第52周流腺报告发病数分别建立指数平滑和季节性自回归滑动平均(seasonal autoregressive integrated moving average,SARIMA)模型,对2种模型比较,对未来流腺疫情预测和预警。结果通过检验的优选模型按拟合效果高低分别为SARIMA(22,0,0)(1,0,1)52模型、ARIMA(1,0,0)模型和指数平滑阻尼趋势模型,评价指标显示ARIMA(1,0,0)模型和指数平滑阻尼趋势模型的拟合效果相差不大,3个模型预测值的95%可信限均覆盖实际值。3个模型对2020年第53周预测值分别为73例、94例和85例,相比2020年第52周走势不同,因模型中季节性因素的有无所致。结论基层或一线工作人员首选便捷的指数平滑模型用于常规性分析,定期建立SARIMA模型对指数平滑分析结果进行评估和修正。

Objective To compare and analyze the incidence of mumps reported weekly in Hebei Province by different time series methods,so as to choose the appropriate method for more convenient prediction.Methods The exponential smoothing and seasonal autoregressive integrated moving average(SARIMA)models were established based on the reported incidence of mumps in Hebei Province from the first week of 2004 to the 52nd week of 2020.The two models were compared to predict and warn the epidemic situation of mumps in the future.Results The best models that passed the test were SARIMA(22,0,0)(1,0,1)52,ARIMA(1,0,0)and exponential smoothing trend model.The evaluation indicators showed that there was no significant difference in fitting effect between the ARIMA(1,0,0)model and the exponential smoothing trend model,and the 95%confidence limits of the predicted values of the three models all covered the actual values.The predicted values of the three models for the 53rd week of 2020 were 73,94 and 85 cases,respectively.The trend was different from that of the 52nd week of 2020,due to the presence or absence of seasonal factors in the models.Conclusion The convenient exponential smoothing model is preferred for routine analysis by grass-root or front-line staff,and SARIMA model is established regularly to evaluate and correct the results of exponential smoothing analysis.