Research On Blood Pressure Prediction Model Of Hypertension Patients Based On Deep Learning

Hypertension is a major problem that harms our public health.Nearly 700,000 people in my country's urban and rural residents die of cardio-cerebrovascular disease each year,more than 40% of the total number of deaths due to disease.Among all deaths,70% of stroke cases and 50% of myocardial infarction cases are closely related to hypertension.In view of the high prevalence and late fatality rate of hypertension,my country has adopted regular follow-up strategies for patients with hypertension.The follow-up physicians of the primary medical and health institutions provide professional antihypertensive guidance for hypertensive patients within the jurisdiction,aiming to improve the quality of life of hypertensive patients.However,the blood pressure level of the human body is complicatedly affected by many factors,either sporadic changes due to the influence of external environmental factors or fundamental changes due to long-term drug treatment or lifestyle intervention,and each influencing factor is reflected in different There are differences in patients.Only based on the follow-up physician's personal medical experience to develop a blood pressure reduction plan for patients,there may be some omissions,and the follow-up management effect of hypertension patients cannot be guaranteed.In response to this problem,based on the follow-up records of hypertension patients in a city of Shandong Province,this article explored the comprehensive effect of various factors such as antihypertensive treatment measures and external environment on the patient's blood pressure level.For the first time,a quantitative output in many aspects was established.The blood pressure prediction model of hypertension patients under the influence of factors.This model can provide strong data support when follow-up physicians develop personalized antihypertensive strategies for hypertensive patients.The main research contents and main achievements of this article are as follows:(1)Collected and processed relevant data on the influencing factors of patients' blood pressure levels.According to the existing survey statistics report and scientific experimental research on the influencing factors of human blood pressure,starting from the three dimensions of the patient's basic characteristics,the way of blood pressure reduction and the external environment,we summarized the gender,age,waist circumference,exercise intensity A total of 26 related factors that affect the patient's blood pressure level,such as medication and medication categories.The collected original follow-up records of hypertension patients and the original weather records are processed to form a characteristic data set of hypertension patients,and the input and output of the blood pressure prediction model of hypertension patients are determined.(2)The three algorithms of support vector machine,back propagation neural network and deep belief network are applied to blood pressure prediction modeling of hypertension patients.Due to the complex influence of many factors on human blood pressure,simple statistical methods are difficult to characterize the comprehensive effect of many factors on the patient's blood pressure,so this paper uses three algorithms with good generalization ability to establish a blood pressure prediction model for patients with hypertension,and The indicators such as prediction accuracy,calculation time and memory occupancy were compared horizontally between models,and the model based on DBN was found to perform best.(3)Improve and optimize the blood pressure prediction model of hypertension patients based on DBN.Combined with the characteristics of the characteristic data set of hypertensive patients,the method of determining the internal unit structure and hidden layer structure of the DBN model is optimized,and the adaptive moment estimation algorithm is used to adjust the parameters in the model.A blood pressure prediction model for hypertensive patients with higher prediction accuracy,greater scalability and faster convergence speed is realized.