Application of error-trend-seasonality model based on state-space in predicting tuberculosis incidence in Henan

  Objective  To evaluate the application of error-trend-seasonality model based on state-space (ETSBSS) in forecasting tuberculosis (TB) incidence in Henan province.

  Methods  Time series decomposition method was used to analyze the trend and seasonal components of the TB incidence in Henan from 2006 to 2019. The data were divided into training set (2006−2018) and testing sets (2019), and then ETSBSS model was used for fitting and prediction, and the model’s fitting and prediction performances were compared with those of the seasonal autoregressive integrated moving average (SARIMA) model.

  Results  The ETSBSS (A, MD, M) and SARIMA (1, 0, 0) (0, 1, 0)₁₂ specifications were selected as the best models to predict the TB incidence in Henan. The mean absolute percentage error (MAPE) values from the ETSBSS model were 5.65% on the training set and 4.61% on the testing set, which were lower than those from the SARIMA model (5.71% on the training set and 6.67% on the testing set). The values of mean absolute error, root mean square error, mean error rate, and root mean square percentage error also indicated that the fitting and prediction error rates of the ETSBSS model was lower than those of the SARIMA model, especially in the prediction set.

  Conclusion  ETSBSS (A, MD, M) model shows a high prediction performance for the TB incidence in Henan, and it can be used as an effective decision-making tool to predict and analyze the dynamical epidemic patterns of TB in Henan.