The early warning model of HFMD which is implemented by the multivariable deep learning neural network

At present, many neural networks have achieved good results in disease prediction. For hand foot mouth disease(HFMD), mastering its epidemic law can provide scientific basis for effective formulation of prevention and control measures. However, most existing models use SIER infectious disease dynamics model, seasonal difference moving ARIMA or RNN, LSTM and other traditional networks to predict, and do not take climate and other factors into account. In this paper, firstly, a new data set is established, and BiLSTM is used to predict hand, mouth and foot disease. In addition, climate factors are taken into account in the prediction process. Thirdly, we conducted the Controlled experiment with the traditional models to calculate their MAE, MSE, RMSE, MAPE and other evaluation values. And the experiment shows that robustness of BiLSTM model in HFMD is better than these models. Finally, we analyzes the model from the perspective of time step, and sets the time step to 7 days, 14 days, and 21 days. It is found that when the time step is 14 days, the prediction performance is the best. Finally, we also made a comparative analysis through ablation experiments, and found that the HFMD dataset with meteorological factors was better than the HFMD dataset without meteorological factors in prediction accuracy.