Classification and Prediction of Students’ GPA Using K-Means Clustering Algorithm to Assist Student Admission Process

Abstract

Background: Student admission at universities aims to select the best candidates who will excel and finish their studies on time. There are many factors to be considered in student admission. To assist the process, an intelligent model is needed to spot the potentially high achieving students, as well as to identify potentially struggling students as early as possible. Objective: This research uses K-means clustering to predict students’ grade point average (GPA) based on students’ profile, such as high school status and location, university entrance test score and English language competence. Methods: Students’ data from class of 2008 to 2017 are used to create two clusters using K-means clustering algorithm. Two centroids from the clusters are used to classify all the data into two groups:  high GPA and low GPA. We use the data from class of 2018 as test data.  The performance of the prediction is measured using accuracy, precision and recall. Results: Based on the analysis, the K-means clustering method is 78.59% accurate among the merit-based-admission students and 94.627% among the regular-admission students. Conclusion: The prediction involving merit-based-admission students has lower predictive accuracy values than that of involving regular-admission students because the clustering model for the merit-based-admission data is K = 3, but for the prediction, the assumption is K = 2.