Explainable Machine Learning Framework for Early Detection of Heart Disease Using Feature Selection and Data Balancing
Though technology is extensively applied to medical field, it remains one of the most urgent problems of healthcare sphere as heart diseases are complicated by the interplay of clinical, behavioral, and physiological causes. This research is aimed at suggesting a machine learning framework that can provide accuracy, transparency, and reliability regarding the prediction of heart disease. During the process of reducing redundancy and computational overhead, the proposed system makes use of sophisticated feature selection techniques in order to make predictions about the risk factors that are the most influential. To reduce the impact of class imbalance and make sure that each category of patients has equal access to education, the SMOTE advanced data balancing strategy is employed. Random Forest, XGBoost and Support Vector Machine are some of the machine learning models that are used to identify the optimal predictive model. The approach is based on explainable artificial intelligence (XAI) methods, which are SHAP values. Such techniques provide an explanation of the choices taken by the model and emphasize the role of each feature in the disease risk. Experimental results on datasets that relate to heart disease, among other things, show the superior performance in accuracy, F1-score, and area under the curve (AUC), as well as a high level of interpretability is maintained. This method of utilization provides a reliable, unbiased, and evidence-based tool that can be helpful to the medical department in the initial risk evaluation of cardiovascular diseases.