Machine Learning Method for Continuous Noninvasive Blood Pressure Detection Based on Random Forest

In order to reduce the influence of differences in human characteristics on blood pressure prediction model and further improve accuracy prediction, this paper establishes support vector machine regression random forest for accurate measurement. First, photoelectric method is used obtain plethysmography signal (PPG) ECG signals from people different ages, value roughly estimated based high-quality physiological vascular elastic cavity model; then body are as input parameters model, find best parameter combination performance finally, through a lot training learning, selected achieve measurement values. It has been verified by experiments that average absolute error diastolic systolic optimization meets standard less than 5mmHg formulated AAMI (American Medical Instrument Promotion Association), which better consistent with mercury sphygmomanometer, more excellent under same conditions.