Prediction Of Heart Failure Patient Survival Using Machine Learning

Heart failure is a common, incurable illness with high morbidity and mortality rates globally. Early detection enables proper medication and monitoring, significantly affecting survival rates, especially in the present healthcare environment driven by the COVID-19 pandemic. Machine learning techniques offer a powerful tool for predicting heart failure survival by uncovering hidden data patterns. This paper aims to predict heart failure patient survival using machine learning, focusing on identifying significant features through techniques like Recursive feature elimination, Random Forest, and Information Gain. The machine learning model used is an ensemble method, a stacked classifier with CatBoost, LightGBM, and XGBoost as base models and Multilayer Perceptron as the meta-learner. The best-performing models were selected from a list of trained models, including Logistic Regression, Random Forest, Support Vector Machine, LightGBM, CatBoost, XGBoost, and Multilayer Perceptron. The heart failure patient survival dataset utilized in this research is artificially created through Gretel AI, a synthetic data generation platform, based on a primary UCI medical dataset (Heart Failure Clinical Records, UCI). This method ensures data confidentiality while preserving essential statistical characteristics. This study contributes to research on predicting heart failure survival by emphasizing early intervention and demonstrating the potential of machine learning in improving patient outcomes.