A Comparative Study of Machine Learning Algorithms for Predicting Cardiovascular Disease

Cardiovascular diseases (CVDs) are one of the primary causes of global morbidity and mortality, presenting significant healthcare challenges. The critical need for accurate prediction of CVD risk is paramount for timely and proactive intervention, yet it remains a challenge. To enhance CVD risk prediction, this study delves into a range of machine learning algorithms, encompassing supervised, and ensemble algorithms. Furthermore, recognizing the literature’s limitations, our focus has been on enhancing model performance through hyperparameter tuning, implementing robust feature selection methods, and conducting thorough model evaluations. Besides, for feature selection, we utilize chi-squared tests and correlation analysis to ensure the relevance and significance of the features. Moreover, our comprehensive evaluation, spanning three diverse datasets, assesses both supervised and ensemble learning algorithms for their accuracy and generalizability. The results indicate that the K-Nearest Neighbors-based model excels, achieving 97.82% accuracy. By enhancing predictive accuracy and model robustness, our study not only contributes to improved patient-specific interventions but also aids in shaping more effective and efficient public health strategies in cardiovascular care.