HeartCare: Web-based Heart Disease Prediction Using Machine Learning

Small public datasets can yield optimistic heart-disease prediction scores when data-dependent preprocessing is fitted before the train-test split. We re-evaluated three models on the combined UCI Heart Disease dataset and deployed the selected model in HeartCare, a Flask web application. After removing one physiologically impossible record, we retained 919 observations and evaluated Random Forest, XGBoost, and SVC on the same 276-record held-out test set with a leak-free pipeline. XGBoost achieved 0.8370 accuracy and 0.9079 ROC AUC; the three models performed comparably. Code: https://github.com/MuhammadSaud60/HEART-DISEASE-PREDICTION-FYP-