Students Performance Prediction in Virtual Learning Environment Using a Deep Learning System

Abstract

These days, virtual learning environments (VLEs) are essential and widely used worldwide for information exchange. While VLE benefits remote learning, in-person lectures are more challenging to maintain student engagement than in-person lectures, contributing to the high dropout rates among students. Student’s learning curves are impacted by their lack of active participation in academic activities. Therefore, the VLE needs to give more attention to severe academic achievement. This paper proposes a novel enhanced activation function-centered recurrent neural network (EARNN) with a K-means Synthetic Minority Over-sampling Technique (KMSMOTE) to predict students’ performance in VLE. The four primary steps of the suggested system are data gathering, preprocessing, balancing, and classification. First, the proposed system collected the data from the Open University Learning Analytics Dataset (OULAD) dataset. Next, the system performs preprocessing on the collected dataset to improve its quality. After that, the data balancing is done using KMSMOTE. Finally, the classification of student performance is done by EARNN, which combines demographic, assessment, and click stream features as input. The outcomes demonstrated that the suggested work performs superior to the existing techniques.