Study On The Incidence Regularity And Prediction Model Of Hepatitis A In Liaoning Province In 2004-2016

ObjectiveIn this study,through descriptive analysis and statistics on a data in Liaoning Province,to understand the epidemiological characteristics and the incidence trend of hepatitis A in Liaoning Province.The incidence of hepatitis A in noncoastal areas establish autoregressive integrated moving average model in Liaoning province(ARIMA model),back propagation neural network model(BPNN model),autoregressive integrated moving average model-reverse propagation neural network model(ARIMA-BPNN model),the optimal prediction model is selected by comparing the prediction effect,investigate the epidemic regularity of disease,to achieve early warning of outbreak and epidemic of hepatitis A,to provide a scientific basis for the prevention and control of hepatitis A.MethodsThe data of hepatitis A incidence from 2004-2016 in Liaoning and other cities come from China disease surveillance information reporting system,and population data from Liaoning statistical yearbook.Incidence of month that the number of monthly cases was divided by the number of population in that year,and the unit was 1/100000.Descriptive analysis was applied to analyze the epidemiological characteristics of hepatitis A in Liaoning Province,including epidemic intensity,time,area,age and occupation.Through a month incidence data to establish ARIMA model,BPNN model and ARIMA-BPNN combined model of noncoastal areas of Liaoning province from 2004-2015,to predict the incidence of hepatitis A in 2016,according to comparing the mean square error between the true value(MSE)and the average relative error(MAPE)of 2016 to evaluate the prediction accuracy of prediction results,the model is the best which the accuracy index is the least.ResultsThe epidemiological characteristics of hepatitis A in Liaoning Province from 2004-2016 showed that the incidence of hepatitis A in Liaoning province first decreased and then increased,and the high incidence time was 2-4 months per year.High risk areas appear in the coastal areas of Liaoning Province.The number of men in the onset of the disease is more obvious than that of women,and both men and women at high risk age are at the age of 25-39.Farmers are still high risk occupations,and the proportion of housework and unemployment is rising.In this study,the monthly incidence rate of hepatitis A in noncoastal areas of Liaoning province as a fitting object,established ARIMA model,BPNN model and ARIMA-BPNN model,the ARIMA model parameters of p=1,d=1,q=1,P=0,D=1,Q=1,S=12,we forecast the incidence of 2016 use ARIMA models(1,1,0)(1,1,1)12,MSE was 0.0060,MAPE was 0.4485;the BPNN model through repeated attempts to determine the network structure for use over the past 3 years the incidence data as the input of the network,under the same period of 1 years the incidence rate of the output of the network,the number of hidden layer neurons is 10 and the network model of BPNN for 3-10-1.We forecast the incidence of 2016 use BPNN,MSE was 0.0109,MAPE was 0.3523;the combination model is the ARIMA model prediction value as the input of the network,the actual value of BPNN model of network output,by constantly trying to build,we forecast the incidence of 2016 use the combination model,MSE was 0.0061,MAPE is 0.4518.By comparing the prediction accuracy,the ARIMA model is selected as the optimal prediction model.ConclusionsThe incidence of hepatitis A in Liaoning is still mainly in the coastal area.The incidence of hepatitis A is mostly in young adults.We should continue to do routine immunization of hepatitis A while carrying out targeted immunization strategy for hepatitis A is very important for prevention and control of hepatitis A.In this study,we get prediction models of noncoastal areas,the actual value in the range of ARIMA(1,1,1)(0,1,1),2 model,we could use ARIMA(1,1,1)(0,1,1)12 model to forecast the future incidence of hepatitis A in noncoastal areas of Liaoning Province,it is the effective tool to control the epidemic of hepatitis A.