期刊名称：International Journal of Computer Science and Network Security
印刷版ISSN：1738-7906
出版年度：2017
卷号：17
期号：4
页码：326-331
出版社：International Journal of Computer Science and Network Security
摘要：In the present paper, three types of anemia including iron deficiency anemia (IDA), ��-thalassemia trait and ��-thalassemia trait (cis and trans) have been investigated. Detection of these three types of anemia is difficult, as their blood characteristics are similar to each other. Also, specialists use some tests to diagnose these disorders that those tests are very time consuming and costly. Thus, providing a model for accurate diagnosing of these kinds of anemia is extremely important. The present study mainly focused on a simple complete blood count (CBC) test instead of using some tests to detect and differentiate between these kinds of anemia in Weka software. In order to suggest an algorithm with the highest accuracy and the lowest mean absolute error, five classification algorithms and a vote algorithm were used and the performance of the vote algorithm was compared with the performance of those five algorithms. The results of this study indicated that combining J48, IBK and Naive Bayes algorithms using voting algorithm with all the features and reduced features had the highest performance with the accuracy of 96.343 and 96.2169, respectively. Using hybrid algorithm (vote) demonstrated that hybrid algorithm increases diagnosis accuracy and decreases error rate in comparison with the single classifiers.
关键词：��-thalassemia trait; ��-thalassemia trait; Iron deficiency anemia (IDA); Vote algorithm; Complete blood count (CBC); Data mining techniques
