Application of wavelet-LSTM model in predicting the incidence trend of pulmonary tuberculosis
Objective To explore the prediction effect of wavelet-LSTM(long short-term memory)model on the incidence of pulmonary tuberculosis(PTB),and to provide reference for the prevention and control of PTB.Methods The data of tuberculosis incidence in China from 2012 to 2018 were decomposed into high-frequency series and low-frequency series through wavelet decomposition,and LSTM model was constructed for the decomposed series.The prediction performance of wavelet-LSTM and LSTM models was verified with the data in 2019.Results When predicting the incidence trend in the next 3 months,the LSTM model had better performance,with a MSE(mean square error),MAE(mean absolute error)and RMSE(root mean square error)of 0.007,0.069 and 0.085,respectively.When predicting the incidence trend in the next 12 months,the wavelet-LSTM model had better performance,with a MSE,MAE and RMSE of 0.046,0.156 and 0.215,respectively.Conclusions The long-term prediction performance of the wavelet-LSTM model was better than that of the LSTM model on the incidence of PTB.The wavelet-LSTM model has certain application value in predicting and analyzing the incidence trend of PTB.
NETL
NSTL
万方数据
