AI CARDIONET- A LIGHTWEIGHT STACKING ENSEMBLE FRAMEWORK FOR EARLY DETECTION OF CARDIOVASCULAR DISEASE

Abstract

Heart disease persists as a primary contributor to mortality worldwide, highlighting the urgent need for faster and more accurate predictive tools in clinical practice. Improvements in diagnostic accuracy could be possible with the help of new AI technologies and supporting cardiology decision-making. This study introduces AI-Cardiologist, a supervised learning framework designed to improve early detection of cardiovascular disease using structured medical records. Because of the problems with using outdated models in terms of scalability and interpretability, a novel lightweight stacking ensemble algorithm, CardioStackNet. The method integrates logistic regression, decision trees, and use of logistic regression as the meta-learner and Gaussian Naïve Bayes as the base-learning algorithm for final predictions. A dataset comprising over 70,000 patient records with features such as cholesterol levels, age, blood pressure, body mass index (BMI), and lifestyle indicators was utilized. Following comprehensive preprocessing and feature engineering, the models underwent training and assessment using stratified train-test splits to ensure balanced performance across classes. Experimental results demonstrated that CardioStackNet achieved a predictive accuracy exceeding 95%, significantly outperforming standalone classifiers about recall and F1-score, while maintaining computational efficiency. The findings underscore the effectiveness of lightweight stacking ensembles in practical diagnostic settings, especially those with limited resources clinical environments. Importantly, the proposed system guarantees not only elevated precision but also interpretability and ease of deployment, rendering it a promising candidate for next-generation AI-powered cardiovascular diagnostics.