Classification of the placement success in the undergraduate placement examination according to decision trees with bagging and boosting methods

Abstract

The purpose of this study is to classify the data set which is created by taking students who placed to universities from 81 provinces, in accordance with Undergraduate Placement Examination between the years 2010-2013 in Turkey, with Bagging and Boosting methods which are Ensemble algorithms. The data set which is used in the study was taken from the archives of Turk-Stat. (Turkish Statistical Institute) and OSYM (Assessment, Selection and Placement Center) and MATLAB statistical software program was used. In order to evaluate Bagging and Boosting classification performances better, the success rates of the students were grouped into two groups. According to this, the provinces that were above the average were coded as 1, and the provinces below the average were coded as 0 and dependent variables were created. The Bagging and Boosting ensemble algorithms were run accordingly. In order to evaluate the prediction abilities of the Bagging and Boosting algorithms, the data set was divided into training and testing. For this purpose, while the data between 2010-2012 yearrs were used as training data, the data of the year 2013 were used as testing data. Accuracy, precision, recall and f-measure were used to demonstrate the performance of the methods in the study. As a result, the performance in consequence of "Bagging" and "Boosting" methods were compared. According to this; it was determined that in all performance measure marginally "Boosting" method produced better results than the "Bagging" method.

Covid-19 Hastalığının Teşhisi için CNN Tabanlı Modeller ile Adaboost Algoritmasının Kombinasyonunun Performans Analizi

“…That means after selecting samples of the first subset randomly, these samples are placed into the original set, then samples of the second subset are randomly chosen again from this original set. In short, each sample has an equal chance to be included in chosen training subset [18]. Then, each subset is trained with the designated classifiers.…”