Prediction of students’ performance in e-learning environment of UTMSPACE program

Abstract

Part-time educational programmes enable workers in both private and public sectors a means of acquiring knowledge and advancing themselves in their career. However, part-time students face some emanating challenges in their studies such as time constraint, inability to see lecturers and utilizing the educational resources due to their work commitments. With the advancement in e-learning technologies, the part-time students are able to empower themselves by interacting with eLearning environment so that the instructor may not be the gatekeeper of education. This dissertation is aimed at predicting the performance of part time students registered in UTMSPACE program based on their interactivity with the eLearning activities in MOODLE and MOOCs, this was achieved with the use of the student log files and some additional data about the particular student. The performance prediction was investigated using Decision Tree (C4.5 algorithm) and Neural Network algorithm techniques, in order to find the best technique for the student’s prediction. Neural Networks out-performed Decision Tree C4.5 algorithms by giving 92% accuracy which was validated using precision and recall analysis of the classifier, while Decision Tree obtained 89.2% accuracy. In addition, the analysis of log files indicates that the rate of interactivity with e-learning environment has a significant impact on their performance as the students with highest interactivity on the MOODLE tend to have higher performance than those with low interactivity rate. From the analysis of the log files we can observe that the students spend more time on e-learning MOODLE than MOOCs, and because of that they are missing advantages of the available resources on MOOCs such as watching lecture videos, participating in quizzes, which may assist them in their study.