ROMJIST Volume 23, No. 3, 2020, pp. 238-249
NJOUD ALANGARI and RAAD ALTURKI Predicting Students Final GPA using 15 Classification Algorithms
ABSTRACT: Predicting students’ performance is one of the important tasks in Educational Data Mining (EDM). Early prediction can help the university to take actions that help students to graduate on time and have the best learning outcomes. It also helps the universities to save billions of dollars spent on students who fail, change major or dropout. In this paper, we report work in student performance prediction using 15 classification algorithms. The experimental results show that we were able to predict students’ final GPA with 91% accuracy for two models Naive Bayes and Hoeffding tree which significantly outperform competitive models across different datasets. The average accuracy for all the 15 classifiers was around 71%. We also analyzed the rules generated by tree-based and rule-based classifiers and found that some courses at early levels can have a major effect on the final GPA. In comparison to other works in the field, we were able to have more comprehensive analysis that produced better accuracy with higher variation of class values. These results show the potential for data mining to improve students’ success rate by informing who are at risk in order for Instructors and administrators to provide the necessary support at the right time.KEYWORDS: Data Mining, Machine Learning, Classification, Prediction, Educational Data Mining.Read full text (pdf)