Predicting Chronic Heart Failure after Myocardial Infarction Using Machine Learning Techniques

Chronic Heart Failure is a major concern in the healthcare systems worldwide as it is prevalent and has adverse effects on patients' health. The traditional methods of diagnosing chronic heart failure have limitations that result in ineffective diagnosis, that leads to poor patient outcomes and increased healthcare costs. This study develops a machine learning model to predict chronic heart failure risk using patients' medical and demographic information. For this study, the Myocardial Infarction dataset was taken from the UCI Machine Learning Repository. The proposed system incorporates methods for dealing with missing data, feature selection and handling class imbalance which are important for accurate prediction in clinical environments. The logistic regression model achieved 74% accuracy with precision, recall, and F1-score of 88%, 77%, and 82%, respectively, and for chronic heart failure instances of 47%, 66%, and 55%, respectively. The SVM model also gave comparable results by achieving 77% accuracy and class-specific metrics of 85%, 84%, and 85% for non-chronic heart failure cases, and 51%, 53%, and 52% for chronic heart failure cases. These results clearly show the possibility of machine learning in early identification of chronic heart failure, thereby reducing clinical time and costs while improving patient outcomes.