PREDICTIVE ANALYTICS IN CARDIOVASCULAR HEALTH: A MACHINE LEARNING APPROACH TO DIAGNOSTIC OPTIMIZATION

Early detection of cardiovascular disease (CVD) is critical for reducing global mortality rates. This study evaluates the comparative efficacy of K-Nearest Neighbors (KNN), Gaussian Naive Bayes, and Random Forest classifiers in a clinical diagnostic setting. A central focus of this research was the optimization of the data preprocessing pipeline. We demonstrate that the selection of imputation strategies significantly alters model performance. Specifically, replacing Median Imputation with Mean Imputation resulted in a critical accuracy improvement from 86.00% to 88.00% in the best-performing model. While KNN showed superior performance during validation, the Naive Bayes model demonstrated the highest robustness on unseen test data, establishing it as the most reliable architecture for this specific domain.