Enhancing Student Performance Prediction with Limited Data in Distance Learning Environments

Abstract

Providing early predictions of student performance assessments is an essential task in the educational system. Previous studies on predicting student performance assessments have traditionally relied on academic scores and test indicators. The utilization of assignments, grades, and exams has been an extensive and successful method for evaluating student performance. However, with the increasing popularity of distance learning, a new perspective has emerged. The Online Learning Management System (OLMS) provides a wide array of features that can be leveraged in various ways to predict student performance. This study aims to propose an alternative approach to predicting student performance assessments by utilizing student engagement in an online learning management system. The study strives to investigate and analyze prospective features based on student activity. Bagging ensemble learning methods are proposed to predict student performance assessments through oversampling datasets. The effectiveness of these prediction models is then compared with various machine-learning models, with the results indicating that the proposed model outperforms others at all comparison levels. Furthermore, the proposed model demonstrates the ability to discriminate and predict student performance assessments based on OLMS-related features.

Keywords: Student Performance Assessment, Ensemble Learning, Machine Learning, Student Performance Prediction