Analysis And Prediction Of Hand, Foot And Mouth Disease Incidence Based On Machine Learnin

Hand,foot and mouth disease(HFMD)is an infectious disease caused by a human enterovirus that affects infants and children under the age of five.HFMD was first reported in New Zealand in 1957.Since then,it has begun to spread around the world.Over the past50 years,HFMD has spread rapidly throughout the Asia-Pacific region and is gradually becoming a heavy burden for countries in that region.China established a national surveillance system for HFMD in May 2008.Over the next decade,China reported an average of more than two million cases of HFMD every year.It has become an increasingly serious public health problem in China.Predicting the incidence of HFMD and analyzing its influential factors are of great significance to its prevention.With the vigorous development of artificial intelligence,prediction models using machine learning methods have shown more superior performance among many infectious disease prediction models,but only a few studies have applied machine learning methods to the prediction of HFMD in China.However,most of these studies only take a certain province as the research object,and the prediction models established by them are not general.Therefore,it is necessary for us to establish a nationwide prediction model for HFMD incidence.It is worth noting that the population density in different provinces of China is not uniform,and the population flux also has strong seasonal characteristics.However,some prediction models only consider the impact of meteorological factors on HFMD incidence,but do not consider the impact of social factors such as population density and population flux,which does not conform to the actual situation of China.The prediction model of HFMD incidence considering both meteorological and social factors needs further study.Besides,most machine learning prediction studies only apply a specific machine learning method,and do not compare the prediction performance of different machine learning models.Therefore,the prediction performance of different machine learning methods on HFMD incidence deserves further study.In this study,the HFMD incidence of 31 provinces in China was taken as the research object,five meteorological factors and two social factors were used as predictors to establish Random Forest(RF)and e Xtreme Gradient Boosting(XGBoost)prediction model.After that,we used the RF model and the XGBoost model to analyze the relationship between HFMD incidence and potential influencing factors in 31 provinces of China,and ranked influencing factors based on their importance.We also applied our prediction models in different regions of China,and compared the performance of them based on the mean square error(MSE)and explained variance score(EVS).In addition,we also studied the delayed effect of population flux on HFMD incidence in different regions of China by reconstructing our input dataset.The main contributions of this study are as follows: 1)We established a prediction model for HFMD incidence in China’s mainland,using meteorological factors and social factors as predictors.2)We used k-means clustering algorithm to divide 31 provinces into four different evaluation regions based on meteorological factors,and evaluates the performance of the two prediction models in different evaluation regions.3)Compared with the RF model,the XGBoost model is more suitable for predicting HFMD incidence in China.4)The results show that meteorological factors and social factors jointly affect the HFMD incidence in China.Among them,average temperature,population density and population flux are the three most significant influencing factors.5)Population flux has different delayed effect in affecting HFMD incidence in different regions.At country level,the duration of the delayed effect of population flux is one month.