Comparative Analysis of KNN and SYM Classifiers for Students’ Academic Performance Prediction

ABSTRACT:

The performance of students in tertiary institutions has been a major indicator for assessing the quality of students and their respective institutions over the years. Placement of students into appropriate department/faculty based on their individual competence would no doubt lead to the production of quality graduates who can contribute substantially to societal development. Recently, a number of studies have attempted to provide a means of predicting the performance of students with the aim of placing them into appropriate departments/faculty programmes.  However, among several earlier proposed methods for students’ performance, k-nearest neighbor (kNN) and support vector machine (SVM) seem promising based on their ease of use and high level of reliability.  Therefore, a kNN and SVM predictive models were developed to evaluate and predict the performance of students in institution of higher learning. A comparative analysis of the performances of these two models were observed using 5-fold, 7-fold, and 10-fold cross validation methods. From the experiment, the best performance was achieved when 7-fold cross validation was used of kNN (69.00%) and SVM (72.24%), which indicates that SVM is a better predictor over kNN. Lastly, our findings show that SVM can help provide a more accurateand robust decision making tool with respect to student performance prediction.

KEYWORDS: University, Student performance, Predictive model, k-nearest neighbor, Support vector machine.