Identification of Hypertension Using AI-Driven Models

Hypertension remains a leading cause of cardiovascular morbidity and mortality worldwide, yet conventional screening methods often lack precision for early risk stratification. This study presents a unified AI-driven framework integrating classical machine learning (ML), deep learning (DL), and hybrid ensemble models for the multiclass classification of hypotension, normotension, and hypertension using routine clinical data. The dataset comprises 612 samples with 151 features per entry, including demographic attributes, physiological measures (e.g., blood pressure, BMI), lifestyle indicators, and medical history. The class distribution is moderately imbalanced, with a higher proportion of hypertensive cases. After comprehensive preprocessing—including normalization, missing data imputation, and clinically informed feature selection—six ML algorithms (e.g., Logistic Regression, Random Forest, XGBoost), six DL architectures (e.g., CNN, RNN, BiLSTM), and multiple hybrid ensembles were evaluated. Among ML models, XGBoost achieved the highest performance with an accuracy of 90%, precision of 94%, and F1-score of 86%. Within DL models, the Convolutional Neural Network (CNN) demonstrated superior accuracy (80%) and F1-score (78%). The best-performing hybrid ensemble-combining Logistic Regression, SVM, Random Forest, XGBoost, and Decision Tree-attained an accuracy of 89%, precision of 92%, recall of 87%, and F1-score of 89%, outperforming all individual models. These findings underscore the effectiveness of ensemble-based learning in enhancing hypertension detection accuracy and support the development of clinically interpretable, data-driven diagnostic tools for early intervention.