Comparison of K Nearest Neighbours And Regression Tree Classifiers Used With Clonal Selection Algorithm To Diagnose Haematological Diseases

Abstract

The aim of this study is to develop a method to improve the classification performance by haematological parameters. In classification problems it has been seen that kNN classifier is often used with the clonal selection algorithm. In this study unlike other studies Gini algorithm is performed instead of kNN classification algorithm and higher success rate is obtained. According to the World Health Organisation’s data nearly 10% of women in the world are anaemia. Anaemia is a disease that disrupts life quality and results in serious effects if not cured. Iron deficiency anaemia is the most common type of anaemia and women suffers this disease comparatively to men. Therefore, in this study, anaemia was preferred as a sample application. It is expected to reach successful results in diagnosis of other diseases by looking at haematological parameters with the proposed method. At the end of the study success ratios of different methods are compared by Receiver Operating Characteristics analysis method. While accuracy in memory-based classification is found as 96%, accuracy in regression tree method classification is 98.73%. Using Gini algorithm instead of kNN a higher success ratio is achieved so CSA surpassed ANN’s success ratio.

Keywords

• artificial neural network
• clonal selection algorithm
• computer aided diagnosis
• k-nearest neighbours
• regression trees.
• bilgisayar destekli tanı
• k en yakın komşu
• klonal seçim algoritması
• regresyon ağaçları
• yapay sinir ağları