| Objective: To establish deep learning models for infectious diseases such as human brucellosis and compare them with traditional time series models to provide a reference basis for formulating corresponding prevention and control strategies for their incidence trends and adopting targeted prevention and control measures in advance.Method: The number of monthly cases of human brucellosis from 2004 to 2019,hand,foot and mouth disease from 2008 to 2019,and influenza from 2004 to 2019 were collected from the Public Health Science Data Center,the National Health Commission,and the Chinese Center for Disease Control and Prevention.The data was divided into training set and validation set,a deep learning long short-term memory(LSTM)network mode was established.The model hyperparameters were adjusted,and then the incidence trend was predicted.The prediction results were compared with the traditional time series autoregressive integrated moving average model(ARIMA)model.The root mean square error(RMSE),the average absolute error(MAE)and the average absolute percentage error(MAPE)were calculated evaluate the accuracy of model prediction.The present study mainly adopted the deep learning long short-term memory network model to predict the incidence of human brucellosis,and in order to further understand the generalization of the deep learning model in the prediction of the incidence of infectious diseases,the incidence of the hand,foot and mouth disease,influenza was also predicted.Both the ARIMA model and LSTM model established in the present study were constructed by R software.Results: Through the observation of the time series chart of the monthly incidence data of human brucellosis,hand-foot-mouth disease,influenza,the results of the test of the stability of the time series,and the data of the three were stable.ARIMA model could be used to analysis and predict.For the human brucellosis data,it was found that the RMSE value of the ARIMA model prediction result was 2430.149,the MAE value was2252.699,the MAPE value was 0.681.The RMSE value of the LSTM model prediction result was 424.472,the MAE value was 342.074,and MAPE value was 0.113.It could be clearly seen from the comparison that the accuracy of the LSTM model prediction was better than that of the ARIMA model.To apply this model to hand,foot and mouth disease,influenza could get similar results.For hand,foot and mouth data,the RMSE value of the ARIMA model prediction result was 96588.470,the MAE value was81562.229,the MAPE value was 1.240,and the RMSE value of the LSTM model prediction result was 41390.361,the MAE value was 33128.254,and the MAPE value was 0.234.Through comparison,it can be seen that the accuracy of the LSTM model for hand,foot and mouth disease prediction was also better than that of the ARIMA model.For influenza data,the RMSE value of the ARIMA model prediction result was219408.951,and MAE value was 94622.873,and the MAPE value was 0.617.The RMSE value of the LSTM model prediction result was 182934.306,the MAE value was60134.950,and the MAPE value was 0.369.It could be seen by comparison that the prediction of influenza the prediction accuracy of the LSTM model was also slightly better than that of the ARIMA model.Conclusion: The present study applied the long short-term memory network model to simulate and predict the incidence of human brucellosis.The prediction results were better than the traditional time series model.It was applied to hand,foot and mouth disease and influenza.The results of LSTM model were better than the traditional ARIMA model.On the one hand,the accuracy of the model was verified.On the other hand,it showed that the long short-term memory network model could be used to predict the incidence trend of infectious diseases.The LSTM model was proved to be generalization.The LSTM model was good at predict time series data,and it was expected to play an important role in predict the incidence of infectious diseases. |
PDF Full Text Request