Children Hematopoietic Stem Cell Transplant Survival Status Prediction using Machine Learning

A surgical treatment known as a bone marrow transplant (BMT) or hematopoietic stem cell transplant (HSCT) successfully treats bone marrow diseases. However, the treatment has a number of risk factors that may reduce long-term survival. This research aimed to predict the survival of children receiving Hematopoietic Stem Cell Transplantation (HSCT) or bone marrow transplant (BMT) using different machine learning classifier models, including Decision Tree (DT), K-Nearest Neighbors (KNN), Random Forest (RF), AdaBoost (AdB), Gradient Boosting Classifier (GBC), and XGBoost (XGB), using a publicly available dataset. The study involved preprocessing the data and balancing it using SMOTEENN to address the class imbalance issue. Feature selection was also performed using chi-square and correlation methods, resulting in the use of 19 out of 39 features. The data was then split into a 70–30 train-test ratio and trained using the aforementioned machine learning classifier models. The results showed that the decision tree classifier had the highest accuracy rate of 96.77%. These findings suggest that machine learning models can be utilized to predict the survival of children undergoing HSCT, with the decision tree classifier being the most accurate. So, this research provides a foundation for future studies that aim to improve the accuracy of survival predictions for children undergoing HSCT. Additionally, the findings may be utilized to aid in treatment decision-making and counseling for patients and their families.