A Machine Learning Approach for Risk Prediction of Cardiovascular Disease

Worldwide, Cardiovascular Diseases (CVD) continue to be the most prevalent cause of fatality and morbidity, claiming the lives of approximately 20.5 million individuals annually. Timely and accurate risk prediction is crucial in identifying high-risk individuals and implementing preventive measures. Leveraging Machine Learning (ML) techniques and unbiased data analysis can enhance the efficacy of risk forecasting by uncovering new risk determinants and understanding complex relationships among them. In this study, publicly available data from the University of California Irvine (UCI) repository was utilized to detect CVD. Four ML methodologies, namely K-Nearest Neighbor (KNN), Multi-Layer Perceptron (MLP), Support Vector Machine (SVM), and Extreme Gradient Boosting (XGBoost), were employed, along with hyperparameter tuning techniques to improve predictive accuracy. The models were trained and evaluated using Grid Search Cross-Validation. The results demonstrated promising predictive capabilities for each classifier. The KNN model achieved an AUC value of 0.9821, the MLP model achieved 0.9935, the SVM had an AUC score of 0.9464, and the XGBoost performed exceptionally well with an AUC score of 0.9978. Consequently, the XGBoost model was recommended for automated CVD detection. This study highlights the potential of ML techniques in the healthcare sector, offering an alternative to conventional visual inspection methods. The widespread adoption of ML technology empowers medical professionals to diagnose CVD more accurately by evaluating complex traits and features, thereby improving patient prognosis and treatment.