Empowering Self-Care: Heart Attack Predictions Using Explainable AI and Machine Learning

As cardiovascular disease continues to be one of the leading causes of mortality worldwide, timely diagnosis is of paramount importance to improve therapeutic outcomes. This research presents an artificial intelligence-based heart attack prediction model using a hybrid approach that integrates Extreme Gradient Boosting (XGBoost) with Deep Neural Networks (DNN) and incorporates SHapley Additive exPlanations (SHap) for improved interpretability. The proposed method reliably predicts an individual’s risk of experiencing a heart attack by using an extensive dataset of various health factors such as age, blood pressure, and cholesterol levels. XGBoost explores non-linear feature interactions, while DNN models analyze more complex patterns within the data. SHAP delineates the contributions of each attribute for individual predictions as well as across the population, providing both local and global interpretability. The proposed model achieves an exceptional accuracy of 98.8%, accompanied by an F 1 score of 96.5%, precision of 95.7%, recall of 96.2%, and specificity of 95.5%. Providing consumers and healthcare professionals with transparent, tailored risk assessments facilitates this innovative approach to promoting early intervention and self-care practices.