Identification of the Tumor Markers in Ovarian Cancer using Different Data Mining Methods

Abstract

Ovarian cancer that continues the most common lethal gynecological cancer is the fourth cause of death from cancer among
women in industrialized countries. Early detection of ovarian cancer is difficult because of typically diagnosed at late stage.
Therefore, early detection and define the identifier has great contribute to improve clinical outcomes. In this study, we searched
the best identifier(s) in early diagnosis as well as the best data mining methods by using ovarian cancer dataset that were
taken from Selcuk University, Faculty of Medicine. The experimental results show that while some identifiers that include
Cancer Antigen 125 (CA125), lesion 1, 2, 3 and mural lesion is the most important identifier as individual basis, combination of
CA125 and lesions are very significant clinical indicators. We can say that CA125 is not considerable identifier by alone and it
should be used with the other identifier although commonly used in the diagnosis of cancer. In addition to this, the classification
tree method achieved the highest success in classifying ovarian cancer data, with a 92.31% success rate in both malign and
benign data.