The Epidemiological Characteristics And The Construction Of Prediction Model On Hand-foot-mouth Disease In China

Objective:To analyze the epidemiological characteristics of hand-foot-mouth disease in China from 2007 to 2016,and to explore the effects of SARIMA model,SVR model,WNN model,ELM model and NAR dynamic neural network in the prediction of HFMD incidence.Through the evaluation of MAE,RMSE and MAPE indicators,the model that best fits the monthly incidence of hand,foot and mouth disease in China is selected.Therefore,it enriched the methodological research of infectious disease prediction model,and broadened new ideas for early warning policy for hand-foot-mouth disease in China.Methods:Through the National Science and Technology Basic Condition Platform,the National Population and Health Science Data Sharing Platform Public Health Science Data Center,and the National Health and Welfare Commission's Disease Prevention and Control Bureau's National Legal Infectious Disease Epidemic Report to obtain HFMD data from2007-2017 in China.To describe its epidemiological characteristics such as time distribution,regional distribution,age distribution and occupational distribution by using Excel 2013 and ArcGIS 10.2 software,and to establish a SARIMA model,SVR model,WNN model,ELM model and NAR model.Simultaneously,the evaluation indicators such as MAE,RMSE and MAPE were applied to evaluate the pros and cons of the model.The smaller the value,the better the model prediction effect.Using EViwes 8.0 software to build SARIMA models,libsvm toolbox of MATLAB 2017a software to build SVR models,MATLAB 2017a software to build WNN and ELM models respectively and GUI toolbox of MATLAB 2017a software to build NAR model.Results:?1?From 2007 to 2016,it was reported a total of 15297683 cases of hand-foot-mouth cases in China,with an average annual incidence rate of 120.9880/100,000;3554deaths,with an average annual mortality rate of 0.02646/100,000.In terms of time distribution,the incidence rate of hand-foot-mouth in China showed a double peak distribution,obvious seasonality and periodicity.The incidence rate was the highest in May-June and the lowest in February.In terms of regional distribution,Hainan Province?373.1988/100,000people?andGuangxiZhuangAutonomousRegion?343.5753/100,000 people?have the highest annual incidence rate.In terms of age distribution,the high incidence of hand-foot-mouth disease in China is 0-5 years old,accounting for 94.2669%.In terms of occupational distribution,the high-incidence population was mainly scattered children,the number of cases was 10,236,345 cases,up to 73.6664%,followed by kindergarten childrens 3,155,061 cases?22.7056%?and student 444,275 cases?3.1972%?.?2?The SARIMA prediction model for the incidence of hand-foot-mouth disease in China is SARIMA?3,1,3??1,1,0?₁₂,and the parameters of the model are statistically significant?P<0.001?.Compared with the actual value,the estimated indicators MAE,RMSE,and MAPE of predicted hand-foot-mouth incidence in 2017 by SARIMA model were 3.5149,4.8827,and 26.1709%,respectively.?3?The SVR model was established for the monthly incidence of hand-foot-mouth disease in China.The input of model was the data of the previous 5 years incidence rate,and the current incidence rate data as the output.The optimal parameters c and parameters g were 11.3137 and 0.0221 respectively.Compared with the actual value,the evaluation indicators MAE,RMSE,MAPE of predicted the incidence of hand-foot-mouth in 2017 using the SVR model were 2.8795,4.0743,24.9287%.?4?The WNN model was established for the monthly incidence of hand-foot-mouth disease in China.The data input of the network structure was the previous 5years incidence rate,the current incidence rate data as the network output.Moreover,the number of hidden layer nodes was 3.Compared with the actual value,the estimated indicators MAE,RMSE,and MAPE of predicted hand-foot-mouth incidence in 2017by the WNN model were 2.4371,3.7155,and 23.5869%,respectively.?5?The monthly incidence rate of hand-foot-mouth disease in China was established as ELM model.The network structure input was the data of the previous 3years incidence rate,and the current incidence rate data as the output.And the number of hidden layer nodes is 15.Compared with the actual value,the estimated indicators MAE,RMSE,and MAPE of predicted hand-foot-mouth incidence in 2017 by the ELM model were 2.1304,2.8234 and 20.9452%,respectively.?6?The NAR model was established for the monthly incidence of hand-foot-mouth disease in China.The LM algorithm was used for network learning.The sample data distribution ratio was 80%for training set,10%for verification set,10%for test set.The number of hidden layer neurons was 8 and delay variables were 12.Compared with the actual value,the estimated indicators MAE,RMSE,and MAPE of predicted hand-foot-mouth incidence in 2017 by the NAR model were 2.5071?3.8756 and 20.9665%,respectively.Conclusion:?1?The incidence of hand-foot-mouth disease in China had obvious seasonal characteristics.The main incidence peaks were concentrated in April-July,especially in May,and the second peak of the disease in September-November,showing a bimodal trend.In 2007-2016,the region with the highest annual average incidence rate in the past 10 years was dominated by Hainan Province and Guangxi Zhuang Autonomous Region.The high incidence of hand-foot-mouth disease in China is 0-5 years old,mainly for scattered children and young children.?2?SARIMA?3,1,3??1,1,0?₁₂ model,SVR model,WNN model,ELM model and NAR model can be used to fit and predict the monthly incidence of HFMD in China.?3?Compared comprehensively,the prediction effect of ELM model is the best than SVR model,WNN model and NAR model in all of the prediction models of monthly incidence of HFMD in China in this study.