Abstract:
Objective
To establish an ARIMA model for inpatients with severe acute respiratory infection in Pediatric Department of Chengguan District People's Hospital of Lanzhou, and to judge the modeling result and make predictions.
Methods
The number of cases of severe acute respiratory infections from January 2010 to June 2016 in the People's Hospital of Chengguan District of Lanzhou City was selected as the modeling part of the ARIMA model. From July to December 2016, the data from July to December 2016 was used as a model validation part to obtain the model's fitting and prediction effect.
Results
From January 2010 to June 2016, the number of severe acute respiratory infections admitted to the Department of Pediatrics of Chengguan District People's Hospital of Lanzhou City was lower in January and February, and higher in June and July, showing a seasonable trend. ARIMA(1,1,2)×(2,1,0)12 model fitting effect was good. The average absolute percentage error of the model verified on the test set was 31.30;the root-mean-square error was 11.49.Taking the data from July to December 2016 as the verification set, the relative error of the verification result was 22.19%.
Conclusion
The established ARIMA(1,1,2)×(2,1,0)12 model has a good fitting effect on the test set. The prediction accuracy on the validation set is high. This model has been applied in the Pediatric Department of our hospital to predict the case number of sever acute respiratory tract infection and to send alarm, which has achieved good results.

Key words: Severe acute respiratory infection, Exponential smoothing method, ARIMA model, Incidence, Prediction, Children