Fig. 4

The proposed model for classification of vasovagal syncope. A flow diagram showing the proposed model for classification of vasovagal syncope. Features ECG and blood pressure signals from 137 HUTT were first extracted then imputed. Selected features identified using feature selection methods and non-parametric probability testing were performed in order to compare the statistical differences between two groups then cross validated and 80% of the data used as training set for machine learning, with the remaining 20% as testing set. The model performance was evaluated, then prediction classification and partial dependence plot applied. HUTT, head-up tilt test; ML, machine learning; SVM, support vector machine; K-nearest neighbors (KNN) imputation dataset; GNB, Gaussian naïve Bayes; MNB, multinomial naïve Bayes; LR, logistic regression; RF, random forest