Blood diseases detection using data mining techniques

Abstract

In healthcare systems, there is huge medical data collected from many medical tests which conducted in many domains. Much research has been done to generate knowledge from medical data by using data mining techniques. However, there still needs to extract hidden information in the medical data, which can help in detecting diseases in the early stage or even before happening. In this study, we apply three data mining classifiers; Decision Tree, Rule Induction, and Naive Bayes, on a test blood dataset which has been collected from Europe Gaza Hospital, Gaza Strip. The classifiers utilize the CBC characteristics to predict information about possible blood diseases in early stage, which may enhance the curing ability. Three experiments are conducted on the test blood dataset, which contains three types of blood diseases; Hematology Adult, Hematology Children and Tumor. The results show that Naive Bayes classifier has the ability to predict the Tumor of blood disease better than the other two classifiers with accuracy of 56%, Rule induction classifier gives better result in predicting Hematology (Adult, Children) with accuracy of (57%–67%) respectively, while Decision Tree has the Lowest accuracy rate for detecting the three types of diseases in our dataset.