Effective Prediction of Cardiovascular Diseases Based on Multiple Clinical Parameters

Abstract

Cardiovascular disease (CVD) remains the leading cause of mortality worldwide, claiming an estimated 17.9 million lives each year according to the World Health Organization . Early detection and accurate risk stratification are critical for preventing major adverse cardiac events and improving patient outcomes. This manuscript presents a comprehensive framework for effective prediction of cardiovascular diseases using machine learning techniques applied to multiple clinical, demographic, and physiological parameters. The proposed system integrates eight classification algorithms—Logistic Regression, Support Vector Machine (SVM), Random Forest, XGBoost, K-Nearest Neighbors, Decision Tree, Gaussian Naive Bayes, and AdaBoost—to predict CVD risk based on eleven clinical features including age, blood pressure, cholesterol levels, chest pain type, and electrocardiogram findings. Utilizing the publicly available UCI Heart Disease dataset comprising 918 patient records, the study employs a robust feature selection approach to identify the five most predictive parameters, reducing clinical data entry burden while maintaining high accuracy. Experimental results demonstrate that the Learning Vector Quantization classifier achieves the highest accuracy of 98.7%, followed by the proposed deep neural network with SelectKBest feature selection achieving 99% accuracy on external validation datasets . The findings indicate that machine learning-based risk engines significantly outperform conventional regression-based models, with ensemble methods and neural networks achieving area under the receiver operating characteristic curve (AUROC) values exceeding 0.90 . This research contributes to the development of accurate, interpretable, and clinically deployable CVD risk assessment tools that enable early intervention and personalized medicine.