Forecast of incidence trend of influenza-like illness by the ARIMA model based on R

Objective To analyze the feasibility of predicting the incidence trend of influenza-like illness (ILI) by using the autoregressive integrated moving average (ARIMA) model, which provided technical support for influenza prevention and control. Methods The daily ILI monitoring data reported by the hospital from the 1st week of 2013 to the 26th week of 2017 were collected, a time series analysis was conducted and a prediction model was established with R. Results The overall consultation rate of influenza like illness (ILI%) monitoring data showed a downward trend and had obvious seasonal character. The best predictive model was ARIMA (0,1,1)(0,1,1)₅₂, of which the residual error Box-Pierce test result was χ²=7.07 (P=0.315), χ²=17.22 (P=0.142). The residual error was a white noise sequence, and the actual values of the prediction results were within the 95% confidence interval (95%CI) of the predicted value. Conclusion The ARIMA model can be available for the prediction of short-term incidence trend of ILI in this hospital.