Predicting Undergraduate Students' Achievement In Ethiopian Higher Learning Institutions By Employing A Machine Learning Approach

Abstract

The goal of this work is to predict student achievement in Ethiopian higher education institutions by employing a machine learning approach. Machine learning offers a variety of tasks that can be used to investigate student achievement. The researcher employed the design science research (DSR) method for conducting the study. The classification task is utilized to evaluate student achievement in this study, and since there are numerous techniques for data categorization, different types of machine-learning methods were used. Information like grade points, grades, sex, previous grade point, previous total grade point, previous total GPA, and other attributes were collected from the student’s management system, and finally, to predict the achievement at the end, major factors affecting the student performance, appropriate machine learning algorithms for predicting student achievement, and interesting patterns in predicting the status of the student achievement were identified. This paper investigated the accuracy of machine learning techniques for predicting student achievement. Using students’ educational datasets, the researcher evaluated the models' performance in terms of accuracy, F-measure, confusion matrix, and receiver operating characteristic curve. Random Forest achieved an accuracy of 100%, k-nearest neighbors achieved an accuracy of 99.5%, logistic regression achieved an accuracy of 98.5%, support vector machines achieved an accuracy of 97%, and Nave Bayes achieved 75%. These results were used to determine how machine learning can be used in making decisions to check students' achievements. This study has attempted to apply machine learning techniques to student data, but future researchers should apply them in other education areas for decision-making and problem-solving concerning the quality of education, student placement, and lecturer evaluation, and design an intelligent system that integrates the discovered classification rules with a knowledge-based system.