Predicting Survivability in Leukemia Patients using Deep Learning

Abstract

Leukemia is the most commonly diagnosed type of cancer in children in the United States. The prediction of its survivability is of great importance for patients as it gives them hope and improves their psychological health. Additionally, it is crucial for physicians in prescribing the proper arrangement of treatment. This research aims to predict the survivability of Leukemia patients using Deep Neural Network (DNN) algorithm. The prediction model of the DNN algorithm is experimentally built using two different methods; cross validation method and ensemble method. The performance evaluation is measured using accuracy, specificity, and sensitivity metrics. For comparison purposes, we use traditional Machine learning (ML) algorithms: Decision Tree, Multilayer Perceptron, Support Vector Machine. The proposed DNN algorithm outperforms the other three ML algorithms in predicting the survivability of a Leukemia patient in both cross validation and ensemble methods.