SARIMA模型与残差自回归模型在甲肝发病率预测中的应用及比较

目的比较季节性求和自回归移动平均(SARIMA)模型及残差自回归模型在甲肝发病率预测中的应用效果。方法根据四川省2004年1月~2009年6月的甲肝月发病率资料,分别拟合SARIMA模型和残差自回归模型,比较两种模型的拟合及预测效果。结果 SARIMA模型的AIC值和BIC值分别为64.98和59.07,残差自回归模型的AIC值和BIC值分别为110.01和103.44;SARIMA模型的拟合值与实际值的MAPE、MAE及MSE分别为0.0349、0.0835及0.0016,残差自回归模型的SARIMA的拟合值与实际值的MAPE、MAE及MSE分别为0.0557、0.1392及0.0050。结论 SARIMA模型的拟合与预测效果优于残差自回归模型。

Application and comparison of SARIMA and Auto-Regressive model in predicting incidence of hepatitis A

OBJECTIVE To compare the efficiency of SARIMA model and Auto-Regressive model in the application of predicting the incidence of seasonal infectious disease. METHODS Monthly incidences data of hepatitis A in Sichuan province from Jan.2004 to Jun.2009 were used to fit SARIMA model and Auto-Regressive model respectively. The incidences of the last six months of 2009 were regarded as the testing data so as to compare the fitting and predictive efficiency of the two models. RESULTS AIC and BIC of SAIMA model were 64.9 and 59.07;AIC and BIC of Auto-Regressive model were 110.01 and 103.44. The MAPE,MAE and MSE of the predicted values by SARIMA were 0.034 9,0.083 5 and 0.001 6,respectively;the MAPE,MAE and MSE of the predicted values by SARIMA were 0.055 7,0.139 2 and 0.005 0,respectively. CONCLUSION The fitting and predicting effects of SARIMA are better than Auto-Regressive model.