ARIMA模型在流感样病例发病预测中的应用

目的建立北京市密云区流感样病例ARIMA预测模型并对所建模型预测效果进行验证和评价。方法对密云区2010-2016年流感监测哨点医院每日报告的流感样病例监测资料建立ARIMA预测模型,并预测2017年流感样病例就诊比例,与实际值比较评估模型预测效果。结果2010-2016年密云区流感样病例监测哨点医院门急诊患者累计5339757例,流感样病例数为47324例,流感样病例百分比平均为0.89%。流感样病例监测数据构建ARIMA(0,1,2)模型,对模型残差进行Box-Ljung检验,P=0.71,残差的自相关系数为白噪声序列,能较好地拟合本数据,模型预测值的动态趋势与实际情况基本一致。建立的模型对2017年数据进行预测,实际值与预测值吻合较好,实际值均在预测值95%可信区间。结论ARIMA模型取得比较满意的预测效果,可以用于流感样病例的短期预测。

Application of ARIMA model in predicting the incidence of influenza-like cases

Objective To establish the Arima Prediction model for influenza-like cases in Miyun district and to verify and evaluate the predictive effect of the model.Methods An ARIMA forecast model was established for the influenza-like case monitoring data reported by the 2010-2016 influenza surveillance sentinel hospital in Miyun district,the proportion of influenzalike cases visits in 2017 was forecasted,and the prediction result were compared with the actual value.Results From 2010 to 2016,the total number of the outpatients was 5339757,among which 47324 were influenza-like cases,the average percentage was 0.89%.An ARIMA(0,1,2)model for influenza-like case monitoring data was constructed,and Box-ljung test was performed on the residual error of the model,and the auto correlation coefficient of residual error was considered as white noise sequence,which could fit the data well.The dynamic trends of the model predicted values are generally consistent with the actual situation.The model predicted the 2017 data.The actual values were in good agreement with the predicted values,and the actual values were all within the 95%confidence interval of the predicted values.Conclusions ARIMA model can be used for short-term prediction of influenza-like cases with satisfactory predictive effect.