Identifying Risk Factors and Survival Prediction of Heart Failure Patients Using Machine Learning

Abstract

Heart failure (HF) is a prevalent and complex clinical syndrome with high mortality, making accurate survival prediction crucial for patient management. In addressing this challenge, our research introduces a method that integrates exploratory data analysis, feature selection, and machine learning (ML) models to identify significant risk factors for HF events accurately. The results highlight important factors that impact HF events including Time, Serum Creatinine, Ejection Fraction, Age, Creatinine Phosphokinase, and Diabetes. Among four ML models, Random forest model stands out for its robustness in predicting HF mortality. This is demonstrated by the performance of model through validation and testing data. Specifically, the performance on the validation set achieves an accuracy of 84.9%, a precision of 81.71%, a recall 71.02%, and an F1-score 74.94%. On testing set, the performance of model achieves an accuracy of 86.67%, a precision of 81.82%, a recall of 69.23%, and an F1-score of 75%.