Study On The Prediction Of Diabetes Risk In China

In recent years,the rapid and healthy development of China’s economy has greatly improved the quality of life of the people.However,the improvement of living conditions has also led to a rapid increase in the number and prevalence of diabetes in "rich disease".According to statistics,China is currently the country with the most diabetes patients in the world.The increasing aging population will also lead to an increasing risk of diabetes in our country,which means that China will face continuous challenges in the prevention and control of diabetes.Therefore,it is urgent to establish an accurate diabetes risk prediction model for the general public.The prediction of the model can help to identify high-risk groups as early as possible and effectively prevent diabetes,thereby reducing the prevalence of diabetes to a certain extent.On the basis of previous studies,this article first evaluated the different risks of diabetes in China from three aspects: illness,death and disability.Then,based on the machine learning algorithm,two types of disease risk prediction models were introduced,BP neural network model and random forest model.In addition,the input variables of the model were screened out in detail,that is,the single-factor and multi-factor analysis of the adult data of 12 provinces in the CHNS database was performed,and 8 variables that were significantly related to diabetes were obtained.Taking these 8 variables as input variables of the model,through continuous learning,two prediction models with optimal structure are established.Finally,based on the five indicators of accuracy,sensitivity,specificity,ROC curve and AUC value,the prediction effects of the two models are comprehensively compared.Based on the above research,we draw the following three conclusions.First,the various risks of diabetes in China have a clear upward trend,and the prevalence of the quantitative risk indicators has the largest increase.Therefore,this article decided to make a prediction study on the risk of disease.Second,During the feature selection stage of the model,we obtained 8factors related to diabetes,namely gender,age,province,household registration,BMI,sleep time,hypertension and myocardial infarction.Among them,age,BMI,sleep time,hypertension and myocardial infarction are risk factors for disease,that is,the older the age,the higher the BMI(the fatter the person),the higher the risk of diabetes;Too short sleep time will affect the body’s metabolism,which will greatly increase the risk of diabetes;and patients with hypertension and myocardial infarction are more susceptible to diabetes.In addition,from the perspective of gender,province,and household registration,the probability of women suffering from diabetes is lower than that of men;the prevalence of diabetes in the central and western regions is lower than that in the eastern regions;and the prevalence in rural areas is lower than in urban areas.Third,In terms of model prediction effects,the random forest model has better prediction effects in all aspects than the BP neural network model.Therefore,the random forest model was finally selected as the best prediction model for this study.