Prediction and diagnosis of leukemia using classification algorithms

Abstract

Algorithms used in data mining techniques are of great importance in the field of health care, especially in the case of getting patterns or models that are undiscovered in databases. In the area of health care, leukemia affects the blood status and can be discovered by using the Blood Cell Counter (CBC). This study aims to predict the leukemia existence by determining the relationships of blood properties and leukemia with gender, age, and health status of patients using data mining techniques. More than 4,000 patients were taken from a blood test laboratory from European Gaza Hospital at Gaza Strip. Three classification algorithms are identified for blood Cancer classification; k-nearest neighbor (k-NN), decision tree (DT) and Support Vector Machine (SVM). These three classifiers were implemented and studied thoroughly in terms of classification accuracy and F-Measure. From our experimental results, it was noticed that the decision-tree algorithm had the highest percentage of 77.30% compared with the other two techniques. In addition, the DT classifier obtains properties regarding outer attributes such as city (eastern regions) that are most vulnerable to leukemia.