A Hybrid Machine Learning Framework with Optimized Feature Selection for Augmenting Survival Prediction via Synthetic Match Generator

Abstract

Heart surgery is the most important thing that can be done for kids with end-stage heart failure, but there is still a big problem with death one year after the transplant. It is very important to get this mortality risk right in order to match donors and recipients more effectively and improve patient results. In this work, we use the ICU heart transplant expiration dataset to estimate the risk of death in pediatric heart transplant patients after one year. We suggest a new method that uses advanced feature selection and group methods to make predictions more accurate. Using Chi-squared tests to pick out key traits and combining multiple classifiers for accurate predictions are part of the method. The results show that the suggested Voting Classifier, which uses both Boosted Decision Tree and ExtraTree models, works very well, as it gets 100% of the votes right. This method is a quick and accurate way to guess the chance of death within a year. It gives doctors useful information for better patient care and finding the best match between recipient and donor in pediatric heart transplants.