CLASSIFICATION OF STUDENTS' ACADEMIC SUCCESS USING ENSEMBLE LEARNING AND ATTRIBUTE SELECTION

Abstract

Students' success in high school plays an important role in shaping their lives, as it also affects their success in university placement. It is very important to be able to predict this situation so that in case of failure, precautions can be taken, and a solution can be produced. If success situations and failure can be predicted, success can be increased and stabilized with encouragement and support. In this study, students' academic performances were tried to be estimated with the datasets prepared with secondary school students in Portugal. The datasets include students' answers about the factors thought to affect their success-failure and their grades. The wide use and efficiency of machine learning algorithms have also affected studies on predicting student success. Different algorithms have been applied using different methods in the datasets and the correct prediction rate was tried to be maximized. Experiments were carried out using the 10-fold cross validation method. Deep learning, multilayer perceptrons, simple logistic regression, decision table, one rule, iterative classifier optimizer, logistic model tree and fuzzy unordered rule induction algorithm have been used to predict the student academic success. These algorithms have been tested with the classical and bagging methods. The experiments also tested the efficiency of the algorithms in predicting student success by selecting features and comparing the results.