A Data Mining Approach for Early Prediction Of Academic Performance of Students

Abstract

Quality Education has been declared as one of the seventeen sustainable development goals of the United Nations [1]. Increased development of technology [2] and access to high-speed Internet at an affordable cost [3] have been major contributors in facilitating a larger audience to obtain an education. When a larger community of students are receiving education, considering mechanisms of early prediction of student performance is important to boost the success of high-performing students and provide extra assistance to low-performing students [4]. Trials have been carried out with the inclusion and exclusion of attributes such as gender of students and medium of instructions at university admission exams, to increase the prediction accuracy of the performance of students. Using up to semester three GPA, the three models: SVM, Naive Bayes, and Decision tree gave accuracies in the range of, 77.0%-80.5%, at predicting the students' overall performance. The marks of the university admission exam, the hometown of the student, and the medium of instructions of the admission exam have contributory roles towards the students' performance.