Predicting Achievement with Artificial Neural Networks: The Case of Anadolu University Open Education System

Abstract

This study aims to predict the final exam scores and pass/fail rates of the students taking the Basic Information Technologies -1 (BIL101U) course in 2014-2015 and 2015-2016 academic years in the Open Education System of Anadolu University, through Artificial Neural Networks (ANN). In this research, data about the demographics, educational background, BIL101U course mid-term, final and success scores of 626,478 students was collected and purged. Data of 195,584 students, obtained after this process was analysed through Multilayer Perception (MLP) and Radial Basis Function (RBF) models. Sixteen different networks attained through the combination of ANN parameters were used to predict the final exam scores and pass/fail rates of the students. As a result of the analyses, it was found out that networks established through MLPs make more exact predictions. In the prediction of the final exam scores, it was determined that there is a low level of correlation between the actual scores and predicted scores. In the analyses for the prediction of pass/fail rates of the students, networks established through MLPs ensured more exact prediction results. Moreover, it was determined that the variables as mid-term exam scores, university entrance scores and secondary school graduation year were of highest importance in explaining the final exam scores and pass/fail rates of the students. It was found out that in the higher institutions serving for Open and Distance Learning, pass/fail state of the students can be predicted through ANN under favour of variables of students which have been found as most the important predictors.