Research On Blood Pressure Prediction Model Construction Based On Machine Learning And Risk Control Strategy

With the change of people’s diet structure,as well as the pressure of life,work and so on,the younger characteristic of blood pressure disease is becoming more and more obvious,and the mortality rate is also increasing.Blood pressure diseases are often accompanied by various concurrent diseases and seriously threaten people’s health.Existing blood pressure prediction models are based on single-factor physiological signals,and do not consider the influence of different postures(sitting,standing and lying)and different exercise states(no exercise,light exercise and heavy exercise)on blood pressure.However,both postures and exercise states affect blood pressure.In view of this,combined with the deficiencies of academic theoretical research,this article used experimental methods to measure the blood pressure data in the three postures of standing,sitting and lying,as well as the three exercise states of no exercise,light exercise,and heavy exercise.This paper took experimentally measured blood pressure data as the research object,excavated the characteristics of blood pressure changes in different postures and different exercises states,and built a multi-parameter model through machine learning methods to predict blood pressure values.Finally,the blood pressure risk rating was constructed on the basis of the prediction model,and the corresponding risk management strategies were proposed,which provided a certain reference for the prediction and risk rating of blood pressure and the management of patients with blood pressure diseases in the future.The main research work and conclusions of this article included the following aspects:(1)Experimental design and feature mining of blood pressure in different postures.The experiment recruited 40 volunteers and measured their blood pressure in sitting,standing,and lying positions,and 120 pieces of experimental data were obtained.Through mining the blood pressure characteristics in different postures,it is found that posture and blood pressure are correlated.There are significant differences in blood pressure between sitting,standing and lying.The blood pressure is highest when standing,followed by sitting,and finally lying.However,the range of changes in blood pressure of different subjects and the range of changes in systolic and diastolic blood pressure of the same subject are different.(2)Experimental design and feature mining of blood pressure in different exercise states.Similarly,the blood pressure values of 40 volunteers in the three exercise states of no exercise,light exercise,and heavy exercise were measured,and 120 experimental data were obtained.Through mining the blood pressure characteristics in different exercise states,it is found that exercise state is correlated with blood pressure.The blood pressure after heavy exercise is higher than that of light exercise and no exercise,and the blood pressure of light exercise and no exercise has little change;the change of systolic blood pressure in different states is greater than that of diastolic blood pressure,and the change of different subjects are different.(3)Constructed a multi-parameter blood pressure prediction model considering different postures and different exercise states based on machine learning methods.The study found that the convolutional neural network model is significantly better than the traditional multiple linear regression model and the BP neural network model,and can achieve higher prediction accuracy.The average prediction deviation of the convolutional neural network model is 3.7817 mm Hg,and the average relative error of the samples is 2.54 %.(4)Proposed risk management and control strategies under different postures and different exercise states.Based on the blood pressure prediction model,this paper carried out blood pressure risk rating,and proposed corresponding risk management and control strategies from the government,society and individual levels.