Research On Infectious Disease Forecasting Technology Based On Deep Learning

Infectious diseases are a class of diseases caused by various pathogens that can be transmitted from person to person,from person to animal or from animal to person.Effective prevention and control of infectious diseases is very necessary,otherwise it can lead to group diseases in a short time,and affect people's health and daily life.For the rapid spread of highly pathogenic infectious diseases,the consequences of finding and compiling infectious disease information from Center for Disease Control workers alone would be fatal.Therefore,it is of great significance to design an infectious disease forecasting model that can accurately predict the future information of infectious diseases for disease prevention,treatment and health decision-making.Aiming at the prediction of the number of people with HFMD in Xiamen city,this paper focuses on the research on the prediction model of infectious diseases based on deep learning from the perspective of time series prediction.The main research work includes the following two points:1.In view of the fact that the prediction variables from different historical time have different importance to the prediction of the number of infectious diseases case in the future,this research proposes an attention-mechanism-based infectious diseases prediction model.The model adopts the encoding-decoding structure with attention mechanism,and both the encoder and the decoder are recurrent neural networks.In the attention-mechanism-based infectious diseases prediction model,the input of different time nodes in the encoder is involved in calculating the weight value of the corresponding hidden state,and the weighted sum of the hidden states in the encoder then is used as the intermediate vector which links the encoder and decoder so that the model can pay attention to the historical information which is helpful to improve the prediction accuracy.The experiment results show that compared with baseline models such as LSTM and ARMA,the attention-based model can further improve the accuracy of the number of infectious diseases case forecasting.2.In view of the complicated relationship between the number of infectious disease case and other variables which is difficult to capture,an infectious disease forecasting model based on recurrent neural network is proposed.Firstly,several recurrent neural networks are used to learn the relationship between each variable and the number of infectious disease case in parallel,andthe complex relationship between all the variables used for forecasting and the number of infectious disease case.Secondly,the fusion network composed of residual network and fully connected network is used to fuse the features acquired by recurrent neural networks to obtain the deeper associations among variables.Finally,the predicted number of future case was obtained.Compared with the baseline models used in this study,the experimental results show that the proposed model based on recurrent neural network can effectively improve the accuracy of the number of infectious diseases case forecasting.To sum up,in order to improve the prediction accuracy of the number of infectious diseases case,the two models of infectious diseases forecasting based on deep learning are proposed in this paper,and their forecasting performance is verified by experiments.