Medical Informatics Algorithmic Research Method And Data Analysis For Diabetes

Changes in the traditional lifestyle of the Chinese population and the rapid transformation of our society have led to the emergence of easy-fat environments,and the environments has made many people face a higher risk of type 2 diabetes,especially the young,the elderly and the people at the bottom of our society.Their special lifestyle and pressure made them face greater risk.Experts believed that the type 2 diabetes has a great relationship with modern lifestyles,unhealthy eating habits,decreased exercise and the obesity.The vast majority of type 2 diabetes has no specific symptoms at an early stage and there are about 9 to 12 years of latent period before the clinical diagnosis appeared.Many patients may be sick for many years but can not be diagnosed,It may occur for many years without being diagnosed and there is a large number of undiagnosed diabetic patients in our national.Therefore,medical and sociological workers should strengthen the study of risk factors for high risk of diabetes.The key to reducing the incidence of diabetes is to screen and prevent high-risk populations.We used 403 real medical data with 16 variables in this subject.First we analyzed the data using R studio and visualized data by scatter plots,histograms,etc.And then we calculated the correlation coefficient between each variable and glycosylated hemoglobin,and the correlation coefficient between each variable and diabetes mellitus.Then we used T test and chi-square test to carry out hypothesis testing.Eventually we got six important variables affecting diabetes: cholesterol,blood glucose,the ratio of cholesterol and high density lipoprotein,high density lipoprotein,age and waist.Next,we simulated the regression model of glycosylated hemoglobin and diabetes mellitus.The correct rate of regression model was 92.8%,the sensitivity was 93.67% and the specificity was 86.05%.Then we libraried the support vector machine package,used 10 fold cross validation to find the optimal parameters.Then we constructed the SVM model with the optimal parameters.The correct rate of the model was 97.67%,the sensitivity was 100% and the specificity was 81.82%.At last,a simple diabetes risk assessment system was designed using C# language and ASP.NET technologies.Patient enters his own body index.The system analyzes the input data,then displays the probability of illness and abnormal physical indicators,thus the high risk of diabetes was screened.The diabetes risk prediction model can easily screen out the high-risk groups of diabetes and provide a practical and effective tool for the prevention and detection of diabetes mellitus.Users can easily use the tool for body detection and monitor their changes in body indicators anytime and anywhere.Screening the people who have no obvious symptoms can popular health education,improve public awareness of diabetes,predicting the risk of diabetes,screening high risk groups with diabetes,improving the detection rate of diabetes,and achieve the goals of early detection,early diagnosis,early intervention.Implementing timely and accurate treatment and intervention measures can greatly delay the progress of diabetes,reduce the incidence of diabetes and its complications,improve their quality of life.