An ARIMA Based Model for Forecasting the Patient Number of Epidemic Disease

Forecasting the number of epidemic disease is very important for CDC (center for disease control and prevention). To improve the forecast accuracy, an ARIMA (autoregressive integrated moving average) based model is proposed in this paper. First, autocorrelation (AC) and partial autocorrelation (PAC) analysis are introduced to establish a stationary time series, where the autocorrelation order, moving average order and difference order are estimated. Secondly, least squares method (LS) is employed to estimate the parameters of the prediction model. Finally, the real data between Jan. and Aug. 2014 coming from a CDC are fed into the proposed model and the forecast accuracy obtained is 92.1 %, which significantly outperforms the simple moving average method currently used in the CDC.