Matching preprocessing methods for improving the prediction of student's graduation

Abstract

the aim of this paper is to improve the effectiveness and efficiency of rule-based learning for predicting student's graduation to help in enhancing the quality of education system by matching two preprocessing methods, which are SMOTE and Releif algorithms. This paper used the real-world dataset, which contains 544 students data, obtained from the registration information system at King Mongkut's University of Technology North Bangkok, Prachinburi Campus, Thailand. This dataset is processed with four rule-based learners (DT, OneR, PART, and DTNB). The experimental results have shown that DTNB is providing improved precision, recall, f-measure, and g-mean compared to other methods. Therefore, DTNB algorithm is used to the significant improvement of the prediction student's graduation. The model obtained from our method is used to plan a program of study that will provide the opportunity to graduate in four years.