Transkriptom Veri Seti Üzerinde Derin Öğrenme Yöntemi ile Klasik Veri Madenciliği Yöntemlerinin Sınıflama Performanslarının Karşılaştırılması

Abstract

In this thesis, Artificial Neural Networks, Random Forest, Support Vector Machines, which are classical data mining methods, and Deep Learning method were used to classify the cancer subtypes. The performances of these methods were compared by using some performance comparison measures like accuracy, Kappa and F measure. For this reason, two different RNA sequencing data sets were used. The first data set is the lung cancer data set which has two classes. It is a balanced data set in terms of class size. The other data set is renal cancer data set. This data set contains three classes and the number of observation in these classes are uneven. Gene sets used in the classification were obtained by using different filters. Therefore the performances of the classification methods in different data sets and filters were examined. For each classification method, specific parameters were optimized and the most appropriate parameters were selected. Deep Learning method which has a deeper structure compared to classical data mining methods, showed a successful performance on the data sets used in this study. In terms of the measurements used in performance comparison, the classical single-layer Artificial Neural Network has lower values compared to other methods.