Building Models for the prediction of Leukemia in Children Using Decision Tree and Neural Networks Techniques

Abstract

The most common type of blood cancer is leukemia and prevalent among children and addition to the lack of medical staff against the increasing number of people with leukemia led to the need to create a model helps in the diagnosis process and facilitate the play by those present in the provision of efficient therapeutic services high and more accurate. In this study, samples of the results of blood tests of the children, which aims to detect any types of leukemia, have been distributed among children in Sudan. samples were taken for a complete blood analysis from a hospital National Center for Radiotherapy and Nuclear Medicine Khartoum and blood Bank ( Radiation & Isotopes Center –Khartoum ) the tests included many types of cancers affecting children. Two types of leukemia were concentrated: Acute Myeloid Leukemia (AML). Acute Lymphoblastic Leukemia (ALL). Ranging from children on one to fifteen years of gender .the results were compared with the results of the annual analytical reports of the hospital's statistics office. the results were quite consistent in determining the increase of acute myeloid leukemia. as this type of leukemia affects cells established in the bone marrow which will later be granular blood cells (Granulocytes) and red blood cells (Erythrocyte).It contains all kinds of cells in the blood. Data mining techniques using the decision tree and neural network algorithms helped to obtain the highest accuracy of 99.43% algorithms by dividing data to 70% for training dataset and 30% for testing dataset.