Prediction Of Blood Pressure Based On Human Physiological Data

Blood pressure has become one of the main threats to human health.Accurate and continuous measurement of blood pressure is the premise of effective prevention and treatment of blood pressure diseases.Traditional cuff-based blood pressure measurement relies on the operation of professional medical staff,and can cause harm to the human body during the measurement process,so it is not suitable for continuous blood pressure measurement.Using machine learning algorithm to model and analyze human physiological data,predicting blood pressure is a feasible way to continuously measure blood pressure,but the traditional blood pressure prediction algorithm has the shortcomings of low accuracy and long training time.In this paper,two methods are proposed to solve the above problems.The main contents are as follows:Firstly,this paper introduces the research status of blood pressure measurement methods and some methods of blood pressure prediction based on machine learning.Blood pressure measurement methods include invasive and non-invasive measurements.Blood pressure prediction methods include clustering,classification and regression.The advantages and disadvantages and application background of the above methods are analyzed.Secondly,the original physiological attribute data such as PPG and ECG are obtained by EIMO sensor.The peak detection method is used to extract the most relevant features such as pulse wave conduction time,heart rate and blood oxygen from the original physiological attribute data.The gradient boosting tree algorithm is used to model the characteristic attribute data and predict the blood pressure of a single person.Thirdly,using CM400 sensors,the above features are obtained by similar methods,and the human body attributes such as height and weight are added to predict the blood pressure of many people through classification and regression tree algorithm.By adding regularization methods such as pruning,limiting tree depth and cross validation methods,the generalization ability of gradient lifting tree and classification regression tree is improved,and the risk of over-fitting is reduced.Finally,the method mentioned above is programmed and implemented.The validity of the method is verified by an example based on real blood pressure data and human physiological attribute data.Gradient boosting tree method is better than the traditional least square method,ridge regression,Lasso regression,ElasticNet,support vector regression and nearest neighbor algorithm in accuracy(±5).The accuracy of predicting low pressure value is over 70%,and the accuracy of predicting high pressure value is over 64%.The prediction time of the classification regression tree algorithm is less than 0.1s,and the accuracy of the high and low pressure values is more than 90%,and the training time of the model is less than 0.5s,which has high timeliness.This paper also puts forward new directions for research methods.