Study On Forcast And Health Management Of Typ 2 Diabetes Mellitus Based On Data Mining

Diabetes is one of the most common chronic noncommunicable diseases.According to the latest statistics of the International Diabetes Federation(IDF),the number of global diabetic patients has reached 425 million.The number of diabetic patients in China is 114 million 400 thousand,ranking first in the world,with a diagnostic rate of only 46.4%.Of these,90% of the diabetic patients were type 2 diabetes.Type 2 diabetes is a life-long disease,which can cause complications,such as heart disease,blood vessels,it not only affects the quality of life of the patients,but also brings a serious financial burden to the patients and the country.Therefore,it is necessary to predict type 2 diabetes early,and to find the high risk population of type 2 diabetes,and establish a health management plan for the population,and finally achieve the goal of controlling the incidence of type 2diabetes.Based on data mining classification technology,parameter optimization technology,classifier evaluation and other related theories,as well as the data of physical examination and the data of type 2 diabetes patients,the type 2 diabetes prediction model is as follows:Firstly,the support vector machine(SVM)prediction model is constructed by using the Weka 3.6.13 software and the grid search algorithm to optimize the parameters of SVM.In order to improve the results of SVM prediction model,best predictor attribute of type 2 diabetes are selected by using attribute reduction,and AdaBoost integrated algorithm is used to combine multiple classifiers.On this basis,the attribute reduction-SVM prediction model and the attribute reduction-AdaBoost-SVM prediction model are constructed respectively.The results of comparative analysis of simulated data show that the attribute reduction and AdaBoost integration algorithm can improve the performance of the prediction model at the same time.Secondly,in order to more efficiently and more accurately search parameters of SVM in a greater scope,the Matlab R2014 a software and the particle swarm optimization algorithm(PSO),genetic algorithm(GA)are be used to optimize the parameters of support vector machine.On this basis,PSO-SVM and GA-SVM prediction model wereconstructed.The results of comparison and analysis of simulated data showed that the PSO-SVM model was more suitable for the prediction of type 2 diabetes.This paper aims to construct a prediction model of type 2 diabetes,which has high accuracy,good performance,and the value of clinical application,the high risk population with type 2 diabetes can be early detected,and the corresponding health management programs are formulated,including food and drink management,sports management,education and psychology managemen,condition monitoring,So as to finish the health management plan to prevent the occurrence of type 2 diabetes mellitus,and reduce the incidence of type 2 diabetes.