Abstract:

Objective To understand the effectiveness of empirical mode decomposition-seasonal autoregressive integrated moving average (EMD-SARIMA) model in predicting pulmonary tuberculosis (TB) incidence trend in the Xinjiang Production and Construction Corps. Methods Monthly reported TB incidence data in the Xinjiang Production and Construction Corps were retrieved and analyzed from January 2010 to December 2024. Data from 2010 to 2023 were included in the training set to establish EMD-SARIMA model, and those from 2024 were used in the validation set to evaluate the predictive performance of the models. Finally, the forecasting performance was compared with that of the single SARIMA model. Results From 2010 to 2024, a total of 26 143 TB cases were reported, with an average annual incidence rate of 58.42 per 100 000 population. The incidence peaked in 2011 (88.30 per 100 000 population) and was the lowest in 2022 (38.19 per 100 000 population), showing an overall downward trend. The original signal was able to be decomposed into five intrinsic mode function (IMF) components and one trend term. The optimal SARIMA model for each component derived from the EMD decomposition was identified based on the lowest Akaike Information Criterion, and the residuals of all these models satisfied the white noise test (all P>0.05). The final prediction results of EMD-SARIMA model were obtained by summing the forecasts of each component. The EMD-SARIMA model achieved lower prediction errors, with its MSE, MAE, RMSE, and MAPE values all being lower than those of the single SARIMA model (0.283 vs. 0.745, 0.473 vs. 0.775, 0.532 vs. 0.863, and 13.587% vs. 21.115%). Conclusion EMD-SARIMA model can more accurately predict the tuberculosis incidence trend in the Xinjiang Production and Construction Corps and has better application value compared to the single SARIMA model.

Key words: Pulmonary tuberculosis, Empirical mode decomposition, Seasonal autoregressive integrated moving average model, Prediction