Classification of replacement success according to decision trees by using bagging and boosting methods

This study was conduted to classify the student those took Undergraduate Placement Exam (LYS) organized by Student Selection and Placement Center (OSYM). For the classification, two ensemble methods, Bagging and Boosting were considered. Therefore, data compromised student exam scores and related variables were retrieved from the archives of Turkish Statistical Institue (TUIK) and Student and Placement Center. The evalution has been done by considering the regulations specified in the Handbook of Student Selection and Placement Center (OSYM). MATLAB and R package programs were used to obtained clustering, corrograms, biclustering and bagging and boosting results. Hierarchical clustering results depicted that the provinces within Eastern Anatolia clustered in the same cluster group. Likewise, the provinces of Marmara, Western Eagen and Southern Mediterranean regions clustered into the same cluster in term of similarities. The provinces of Center Aanatolia and Center and Eastern Black Sea, as well as Gaziantep and Elazığ, clustered into the same cluster in terms of similarities distance was used for clustering. These classifications were supported by bagging and boosting. Frther, the conclusion has been done that bagging and boosting have a good predictive ability to classify results by decrease the sum of the squared error term. Keywords:AdaBoost, Bagging, Boosting, Bootstrapping,Ensemble Models, Clustring.