Heart Failure Patient Survival Prediction Using Machine Learning

Abstract

This research presents Heart Guard AI, a machine learning-based clinical decision support system for predicting heart failure patient survival outcomes. The system employs an XG Boost classifier trained on 5,000 clinical records with 12 physiological features. SMOTE is applied to address class imbalance, and Select K Best with Chi-square test performs feature selection. Benchmarked against five algorithms, the system achieves 99.70% accuracy and AUC-ROC of 0.9998. A Flask web application provides real-time High Risk / Low Risk predictions with factor-level clinical analysis.