Performance improvement for diagnosis of colon cancer by using ensemble classification methods

It is an important issue reducing the number of features in a data set to improve the performance of classification algorithms. This effort can also reduce the need of computation power. There are many feature reduction methods used for this aim. In this study, two different feature selection methods have been implemented to the Colon Cancer data set by using the WEKA Data Mining Program. In addition, algorithms using ensemble methods can further improve the classification performance. In this study, different ensemble methods have been implemented to the data set having reduced features. Performance improvements obtained by this way have been evaluated referring to the individual and ensemble classification methods. Performance of the classification methods have been compared by using the Kappa, Accuracy and MCC values. Effectiveness has been shown with ROC graphics. The results have shown that the classification accuracy for the colon cancer can be increased by reducing the features and by using the ensemble methods.