Researchon Decision Support Systemfor Hypertension Management

The unfavorable self-government situation and the failure of effective control of blood pressure of hypertensive patients will give rise to a series of acute and chronic complications.Insufficient communication between doctors and patients will lead to the failure of timely mastering of progression of disease by doctors,resulting in failure of individualized and accurate medical treatment guiding activities by doctors.It has been a problem demanding prompt solution to help hypertensive patients improve self-government level and form preferable interaction mode between doctors and patients with information means.Based on Android platform,Visual Studio 2013 tool and MySQL database design,this thesis realizes the hypertension-oriented chronic disease management aid decision making system,in which WCF framework is adopted for the server-end.The system is divided into the patient-end and the doctor-end;in which the patient-end realizes functions of collection and supervision of physical sign data such as blood pressure and heart rate,behavior supervision,chronic disease risk evaluation and pre-alarm of hypertension;the doctor-end realizes the function of auxiliary health guidance to patients by doctors.The contents of the thesis mainly include three modules of chronic disease risk factor extraction of hypertension,chronic disease classified diagnosis of hypertension and chronic disease risk evaluation of hypertension,with the following specific research contents:1.The mutual information,genetic algorithm(GA)and Bayes Na?ve(BN)algorithm are introduced to extract the characteristics of the chronic disease risk factors of hypertension;the mutual information provides the genetic algorithm with preferable starting point of searching;the GA roulette wheel selects the strategy;the BN algorithm selects the optimal characteristic subset as the evaluation function;the crossing method is adopted for verification.Comparisons are made with the best first search(BFS)algorithm and the sequential floating forward selection(SFFS).According to the characteristic selection,the classification accuracies are 87.50%,83.92% and 85.71%,respectively.2.The weighed maximum voting based classifier fusion algorithm is introduced,to conduct classifier training based on four single classification algorithms of support vector machine(SVM),k-nearest neighbor(KNN),Bayes Na?ve(BN)and back-propagation neural networks(BPNN).Conduct weighing on different classifiers through classifier weighing and classified ranking,and the prediction accuracy of the weighed maximum voting based classified algorithm can be increased by more than 5%.3.The least square method is introduced to construct the 1-1 model,to calculate the relationship between the single risk factor and the chronic disease of hypertension;the Bayes Na?ve algorithm is introduced to construct the n-1 model,to evaluate the relationship between multiple factors and the chronic disease of hypertension.In this way,it extracts the factors of strong relevancy with the chronic disease of hypertension for disease diagnosis and risk evaluation,to provide assistance for doctors to conduct analysis and treatment on the chronic disease of hypertension.