The Study Of Classification Algorithm Of The Type ? And ? Diabetes Based On Wavelet Analysis And C-SVC

Diabetes is a chronic non-infectious disease that seriously harms human health,it can be divided into four categories: type ?,type ?,specificity and gestational diabetes.At present,the classification of diabetes is mainly based on the patient's clinical manifestations,medical examination results and combining with their own experience of diagnosis.It is sometimes difficult to accurately distinguish type ? and type ? diabetes in clinical practice.The main work of this paper is to collect sequences of the patients' blood glucose concentration value through changing time through the Continuous Glucose Monitoring System(CGMS),and to use artificial intelligence to classify patients that provide support for the doctor's clinical diagnosis.For type ? and type ? diabetes,I used CGMS to collect blood glucose data from 1200 patients,at Department of Endocrinology,People's Hospital of Henan Province.Wavelet analysis was used to extract the characteristics of blood glucose curve signals,and the secondary vector C-SVC(Support Vector Machine)was used,SVM)model,the classification algorithm of type ? and type ? diabetes was established,and a new index of type ? and type ? diabetes was established.The main work of this paper includes:(1)The collected blood glucose data was preprocessed,with multi-scale and different wavelet functions analyzing.Using 24 different wavelets,Harr,dbN,symN,coifN and dmey for wavelet decomposion of 2-5 layers,and the wavelet maximum,minimum,median,average and wavelet average energy of each decomposition layer were used as Eigen value.(2)A classification algorithm for type ? and type ? diabetes based on C-SVC model was proposed.The effects of linear,polynomial,RBF and sigmoid kernel functions in C-SVC model are analyzed.The parameter values of C-SVC model are calculated by grid parameter method and genetic algorithm respectively,and appropriate parameter values are selected to improve coincidence rate between the classification results and the clinical diagnosis of C-SVC.(3)The classification algorithm was tested using the collected 1200 patients' blood glucose data.Experiments show that the maximum,minimum,mean and median of the wavelet coefficients after coif4 five-layer decomposition are selected as the best Eigen value,and the classification results of the proposed classification algorithms of type ? and type ? diabetes comparing with Clinical diagnosis has 92%~ 98% coincidence rate.Lastly I compared it with the existing diabetes classification algorithm based on AdaBoost model.