Predicting the Survival Status of Patient after Bone Marrow Transplant Using Fuzzy Discernibility Matrix

Bone Marrow Transplant (BMT), also known as Stem Cell Transplant, is a type of medical treatment in which bone marrow having unhealthy cells is replaced by healthy cells. Bone Marrow Transplant helps to cure various types of cancer like lymphoma, leukemia, and other diseases that affect the bone marrow. Being able to predict if a patient can withstand a Bone Marrow Transplant can be very advantageous for doctors and patients as this prediction can be life-saving. In this paper, we have predicted the survival status of patients after Bone Marrow Transplant with the help of machine learning classifiers with the help of two feature selection processes known as the Fuzzy Discernibility Matrix (FDM) and Principal Component Analysis (PCA). We have calculated the Accuracy score, Precision Score, Recall Score, and ROC-AUC score for every classification model, and a confusion matrix is also plotted for a better analysis of the result. We have also highlighted the comparison of results given by machine learning classifiers using two approaches, where the first approach considers every feature after cleaning the dataset for training and testing, and in the second approach, we have performed feature selection by implementing Principal Component Analysis (PCA) and Fuzzy Discernibility Matrix (FDM). The best results were given by ADA Boost when trained using the features selected by the process of the Fuzzy Discernibility Matrix. It achieved an accuracy score of 0.9523, 1.0 precision score, 0.9259 recall score, and 0.9444 ROC-AUC scores. ADA Boost along with FDM had the least amount of training (64689 $\mu$s) and testing time (5021 $\mu$s) when compared with other approaches. It can also be concluded that using the FDM approach for feature selection will give better results as compared to other traditional feature selection techniques.