应用ARIMA-BPNN组合模型预测手足口病发病率

比较ARIMA模型、BPNN模型和ARIMA-BPNN组合模型在我国手足口病月发病率预测中的应用效果，探讨预测手足口病发病率的优化模型。方法 收集2008年5月至2014年12月我国手足口病月发病率资料，用Excel2007、Spss17.0、Eviews8.0和Matlab8.5拟合相应的模型，并用2014年7月 - 12月的数据评价模型的预测效果。结果 ARIMA模型、BPNN模型和ARIMA-BPNN组合模型拟合及预测的MRD，MSE，RMSE和MAE分别为19.072、2.655、 2.852、 0.182和16.483、 7.246、 2.692、 0.166； 9.031、 2.559、 1.600、 0.009和10.210、 2.655、 1.629、 0.103； 7.397、1.207，1.099、0.068和7.847、1.234、1.111、0.074。结论 ARIMA-BPNN组合模型拟合及预测效果优于ARIMA模型和BPNN模型。

Forecasting Incidence of Hand-Foot-Mouth Disease with the ARIMA-BPNN Combination Model

Objective The aim of this study was to compare the effect of ARIMA model, BPNN model and ARIMA-BPNN combination model in prediction on the incidence of hand-foot-mouth disease in China, so as to explore the optimized model. Methods The data of monthly incidence of hand-foot-mouth disease from May 2008 to December 2014 in China was collected and Excel2007, Spss17.0, Eviews8.0 and Matlab8.5 were used to fit responding models. At the same time, the monthly data from July to December in 2014 was used to evaluate the effect of prediction. Results The MRD, MSE, RMSE and MAE fitted and predicted by ARIMA model, BPNN model and ARIMA-BPNN combination model were 19.072, 2.655, 2.852, 0.182 and 16.483， 7.246， 2.692， 0.166； 9.031， 2.559， 1.600， 0.009 and 10.210， 2.655， 1.629， 0.103; 7.397， 1.207， 1.099， 0.068 and 7.847, 1.234， 1.111， 0.074, respectively. Conclusion According to the model fitness and prediction accuracy, the ARIMA-BPNN combination model with a better practical value was superior to the single ARIMA model and BPNN model.