Research On Early Warning Model Of Influenza In Children Based On Clinical Data Center

So far,the World Health Organization（who）has announced a total of six international public health emergencies（pheic）,such as new crown pneumonia and Ebola.The outbreak of influenza A（H1N1）（a subtype of influenza）in 2009 is one of the six emergencies.According to who reports,about 1billion people worldwide are infected with influenza every year,of which about 10% are adults and 25% are children,causing about 350000 deaths.Because children have weak resistance and are prone to complications,they are often considered as high-risk groups of influenza.The occurrence of influenza in children is caused by many factors,which may be either accidental changes due to the influence of external environment or the inevitable result of personal factors.In order to timely understand the epidemic dynamics of influenza in children and make scientific prevention and control measures for the epidemic of influenza in children.Based on the data of influenza patients collected from the clinical data center of a children’s Hospital in Qingdao,this paper established a prediction model of influenza in children under the influence of various factors.To explore the development trend of children’s influenza in the future and provide data support for hospitals to predict the development trend of children’s influenza.The main contents of this paper are as follows:1)The periodic parameters of ARIMA model are optimized.Because the prediction of influenza in children has the characteristics of periodicity,and ARIMA has the advantage of capturing time series data with periodic changes,however,there is no specific parameter for the selection of ARIMA model cycle.The accuracy of ARIMA model under different cycles is calculated by the least square method,and the optimal value is taken as the input parameter of ARIMA model.Experiments show that the predicted value of ARIMA model improved according to the periodic characteristics is consistent with the actual data,It can more accurately predict the peak of influenza in children in the future.2)The over fitting problem of deep learning LSTM model is improved.Although ARIMA model can predict the non-linear development trend of children’s influenza,in essence,ARIMA model converts non-linear data into linear data through difference for prediction.ARIMA model has good advantages in short-term prediction.LSTM model is mainly used for long-term prediction and processing non-linear data.In the deep learning LSTM model,due to the over fitting phenomenon of the LSTM model,that is,because there are too many indicators and the model is too strict in training,the accuracy of the LSTM model in predicting children’s influenza is low.Therefore,the dropout algorithm is introduced into the model training process to complete the LSTM model training,improve the generalization ability of the model and solve the problem of LSTM over fitting.The experimental results show that the accuracy of the improved LSTM model in predicting future influenza in children is close to 85%.3)Build arima-lstm combination model.ARIMA model and LSTM model have achieved good results in the prediction of influenza in children.However,the single model prediction is often one-sided.Therefore,through linear regression modeling of ARIMA model and LSTM model,this paper constructs arima-lstm combined model and applies it to children’s influenza prediction.Comparative experiments show that the combined model has higher accuracy than the single model,and arima-lstm model has the highest accuracy,nearly 90%.4)This paper also established a visual system for the prediction of influenza in children.Through the establishment of thermal map,line chart,histogram and other visual interfaces,it clearly shows the multi angle influenza prediction of children’s patient age,virus type and so on.The prediction model of children’s influenza studied in this paper has high accuracy and reliability.It can well predict the development trend of children’s influenza in the future,understand the epidemic dynamics of children’s influenza,improve the efficiency of children’s influenza prevention and control,reduce the financial burden of government medical care and personal disease economic burden,provide scientific basis for the prevention and control of children’s influenza,and have important practical significance for the disease prevention and control and disease diagnosis of the whole children’s influenza.