Predicting Study Programme Selection with Data Mining Classification Technique

Abstract

The application of data mining in the field of education (Educational Data Mining - EDM) is becoming more and more popular. Predicting final grades during studies, measuring student and lecturer performance, targeting students, curriculum improvement, are just some of the examples that can support the development of this area. The focus of this work is on the prediction of the study programme that students will select during their higher education at the Faculty of Business Economics in Bijeljina. The analysis was conducted on the data of the faculty wherein the first two years students attend the same courses, while at the beginning of the third year they select a specific study programme. The aim of this paper is to use classification methods to predict the selected study programme based on the final grades achieved on courses during the first two years of study. The highest accuracy was obtained using random forest algorithm (59,94%). Model evaluation results show that choice of study programme does not depend only on the success achieved in all courses during the first two years of study. The analysis was performed using open source WEKA mining tool, and the obtained results were presented and interpreted.