Study On Evaluation Methods Of Human Glucose Tolerance

Diabetes is the third largest chronic non-infectious disease in the world.The number of people with diabetes is increasing year by year,which brings heavy burden to society and families.Because there is no obvious disease performance in prediabetes,it leads to a large number of people cannot find their own health status in time,even cannot realize that they have become a member of the diabetes army.In order to reduce the number of diabetic patients,improve the accuracy of prediabetes screening,effectively evaluate the level of glucose tolerance,establishing a more accurate diabetes screening model is more important.In this study,blood glucose was measured by a noninvasive method.The data set of this paper consists of volunteer physiological parameters,glucose tolerance experimental data and fasting insulin value.On the one hand,the data were analyzed and processed,and the higher risk factors for diabetes were taken as input parameters.The data are classified by KNN and Support Vector Machine and BP neural network,in order to establish glucose tolerance classification models with higher accuracy,which played an auxiliary role in early clinical screening for diabetes.In addition,an easy-to-operate diabetes auxiliary detection model is established in MATLAB.On the other hand,the k-means algorithm and fuzzy clustering are used to cluster all sample sets to establish glucose tolerance clustering models and evaluate the level of glucose tolerance.Provide a new research idea for unlabeled diabetes data.When glycated hemoglobin is added to the input characteristics,the accuracy rates of SVM and BP neural network are 93.46% and 94.77% respectively,which is 0.94%and 3.49% higher than the original model.These two classification models can effectively evaluate the level of glucose tolerance and play an auxiliary role in the detection of diabetes.The research plan of this paper is proposed based on the basis of diabetes screening models at home and abroad and the application of machine learning algorithms in medical data.The classification and clustering algorithms are applied to diabetes data.Analyze the higher risk factors for diabetes and create an early detection model for diabetes.Cluster all sample sets and glucose tolerance assessment models are created.Provide a new research idea to assess the level of glucose tolerance for unlabeled data in the diabetes database.