A hybrid model for tuberculosis forecasting based on empirical mode decomposition in China

BMC Infectious Diseases

Research

Jing Liu¹  Ruiqing Zhao¹  Lixia Qiu¹  Yu Cui¹  Xuchun Wang¹  Yiting Li¹  Mengmeng Zhai¹  Yuchao Qiao¹  Hao Ren¹  Jiahui Ren¹  Zhiyang Zhao²  Limin Chen³ 

[1] Department of Health Statistics, School of Public Health, Shanxi Medical University, Taiyuan, Shanxi, China;School of Public Health, Sun Yat-Sen University, Guangzhou, China;Shanxi Provincial Peoples Hospital, Taiyuan City, Shanxi Province, China;

关键词: Forecasting;    Empirical mode decomposition;    Hybrid model;    Time series analysis;    Tuberculosis;   

DOI  :  10.1186/s12879-023-08609-x

 received in 2023-03-01, accepted in 2023-09-13,  发布年份 2023

来源: Springer

【 摘 要 】

BackgroundPulmonary Tuberculosis is a major public health problem endangering people's health, a scientifically accurate predictive model is of great practical significance for the prevention and treatment of pulmonary tuberculosis.MethodsThe reported incidence data of pulmonary tuberculosis were from the National Public Health Science Data Center (https://www.phsciencedata.cn/). The ARIMA, LSTM, EMD-SARIMA, EMD-LSTM, EMD-ARMA-LSTM models were established using the reported monthly incidence of tuberculosis reported in China from January 2008 to December 2018. The MSE, MAE, RMSE and MAPE were used to evaluate the performance of the models to determine the best model.ResultsComparing decomposition-based single model with undecomposed single model, it was found that: when predicting the incidence trend in the next year, compared with SARIMA model, the MSE, MAE, RMSE and MAPE of EMD-SARIMA decreased by 39.3%, 19.0%, 22.1% and 19.8%, respectively. The MSE, MAE, RMSE and MAPE of EMD-LSTM were reduced by 40.5%, 12.8%, 22.9% and 12.7%, respectively, compared with the LSTM model; Comparing the decomposition-based hybrid model with the decomposition-based single model, it was found that: when predicting the incidence trend in the next year, compared with EMD-SARIMA model, the MSE, MAE, RMSE and MAPE of EMD-ARMA-LSTM model decreased by 21.7%, 10.6%, 11.5% and 11.2%, respectively. The MSE, MAE, RMSE and MAPE of EMD-ARMA-LSTM were reduced by 16.7%, 9.6%, 8.7% and 12.3%, respectively, compared with EMD-LSTM model. Furthermore, the performance of the model were consistent when predicting the incidence trend in the next 3 months, 6 months and 9 months.ConclusionThe prediction performance of the decomposition-based single model is better than that of the undecomposed single model, and the prediction performance of the combined model using the advantages of different models is better than that of the decomposition-based single model, so the EMD-ARMA-LSTM combination model can improve the prediction accuracy better than other models, which can provide a theoretical basis for predicting the epidemic trend of pulmonary tuberculosis and formulating prevention and control policies.

【 授权许可】

CC BY   
© BioMed Central Ltd., part of Springer Nature 2023

【 图 表 】

Fig. 5

Fig. 1

Fig. 2

12951_2015_155_Article_IEq47.gif

Fig. 4

Fig. 5

Fig. 7

Fig. 3

Fig. 3

Fig. 6

Fig. 2

Fig. 3

Fig. 4

Fig. 5

Fig. 3

12951_2015_155_Article_IEq44.gif

Fig. 3

Fig. 1

Fig. 1

Fig. 4

Fig. 4

Fig. 2

Fig. 1

Fig. 3

Fig. 2

Fig. 2