Blood Cancer Prediction using Leukemia Microarray Gene Data and Deep Learning

Abstract

The diagnosis of blood cancer with the use of any leukemia microarray gene se quence data and a machine learning approach is one of the most important fields of medical research. More advancements are needed to obtain the requisite accu racy and efficiency notwithstanding research efforts. Our work’s major goal is to present a method that, using microarray gene data, can accurately predict blood cancer. By increasing the classification accuracy for automated analysis of microarray data analysis, our research seeks to suggest a deep learning model to identify and categorize various types of leukemia. We will use the Leukemia GSE28497 dataset for training our model which contains 281 samples consisting of 22,285 genes (features) of 7 target classes. We preprocess the dataset by deleting null items before training our models. For the prediction of the blood cancer classes, we investigate three classification algorithms: logistic regression, single-layer neural networks, and TabNet. We use a variety of met rics, such as model accuracy, model loss, confusion matrix, train value accuracy, train value loss, and ROC curve, to measure the performance of our models. The outcomes of our studies analyze the effectiveness of deep learning models for clas sifying different forms of blood cancer from microarray gene dat