Predicting Heart Disease using Machine Learning Approach

Heart disease remains one of the leading causes of death worldwide, making early diagnosis and intervention essential. This study explores the application of machine learning algorithms to accurately classify patients with heart disease. To enhance prediction accuracy, we developed a stacked ensemble model that integrates multiple classifiers, including XGBoost, Random Forest, Support Vector Machine, and Gradient Boosting. The dataset was preprocessed by addressing missing values, normalizing features, and balancing class distribution to improve model performance. We evaluated the models using standard classification metrics such as log loss, ROC-AUC, specificity, F1-score, accuracy, precision, and recall. Among the individual models, Random Forest demonstrated the highest predictive performance. However, the ensemble model surpassed all with a ROC-AUC score of $89.56 \%$ and an accuracy of $89.78 \%$. Feature importance analysis identified key parameters-such as age, ST depression, cholesterol, maximum heart rate, and exercise-induced angina-as significant predictors of heart disease. ROC and precisionrecall curves further validated the robustness of the ensemble model. The results underscore the potential of stacked ensemble learning in medical diagnosis, offering clinicians a reliable decision-support tool. Future work may focus on expanding the dataset, incorporating deep learning models, and refining feature engineering strategies to further improve predictive accuracy. These findings support the ongoing advancement of machine learning in healthcare, particularly in cardiovascular disease prediction.