<![CDATA[This study formulates a Support Vector Machine (SVM)-based hypertension classification model using physiological features from Photoplethysmograph (PPG) signals and systolicdiastolic blood pressure estimates. PPG data were collected from 276 participants in three locations using a non-invasive device integrated into the MR-IAT Robot Covid web platform. Feature extraction and preprocessing were performed in MATLAB, while SVM (scikit-learn) model training and testing were conducted in Google Colab. Comparison of three kernel variants—linear, quadratic, and cubic—showed that the cubic kernel was the most superior with an accuracy of 96.4%, followed by quadratic at 94.9% and linear at 91.3%. Overall, the model achieved 93.9% accuracy in distinguishing six blood pressure categories (Hypotension, Normal, Prehypertension, Stage 1 Hypertension, Stage 2 Hypertension, Crisis). Visualization of the results (scatter plot, confusion matrix, parallel coordinates) revealed that systolic pressure, diastolic pressure, age, weight, and respiration rate were the most influential parameters. These findings underscore the potential application of SVM in portable devices for early detection and real-time monitoring of hypertension.]]>
Copyrights © 2025
