PREDICTING HEART DISEASES USING MACHINE LEARNING AND DIFFERENT CLASSIFICATION TECHNIQUES

Abstract

Cardiovascular disease remains one of the leading causes of mortality worldwide, demanding intelligent and accessible diagnostic support systems for early risk identification. This study proposes an enhanced machine learning–driven framework for heart disease prediction by integrating heterogeneous clinical datasets and advanced preprocessing strategies. A combined dataset constructed from a publicly available benchmark source and a privately collected clinical dataset was utilized to improve generalizability and robustness. Data imbalance was addressed using Synthetic Minority Oversampling Technique (SMOTE), ensuring fair representation of high-risk patients. To identify the most informative attributes, three statistical feature selection approaches—Analysis of Variance (ANOVA), Chi-square testing, and Mutual Information—were employed to construct optimized feature subsets. Ten supervised machine learning classifiers were systematically evaluated, and their performance was compared using accuracy, sensitivity, specificity, precision, F1-score, and Area Under the ROC Curve (AUC). Among the evaluated models, the optimized XGBoost classifier demonstrated superior predictive capability. Furthermore, Explainable Artificial Intelligence (XAI) using SHAP analysis was incorporated to interpret feature contributions, enhancing transparency and clinical trust. The final optimized model was deployed within a mobile-based application to provide real-time heart disease risk assessment. The proposed framework offers a reliable, interpretable, and scalable solution for early cardiovascular disease detection