Predicting Students'Final Exam Grades Based on Learning Material Usage extracted from Moodle Logs

Abstract

This paper presents the use of data mining to predict students' final exam grades. We used the data collected from the Moodle platform of the IT course (the University of Split) and compared six classification methods: Decision Tree Classification., k-Nearest Neighbor Classifier., Logistic Regression., Naive Bayes., Random Forest., and Support Vector Machine. Using those methods and Moodle Logs., we aimed to predict the ultimate success in the chosen course. To achieve better accuracy., we evaluated all available and filtered data to determine which algorithms were the most accurate.