Author: Adam James Wilden
Wilden, Adam James, 2025 Predictive Data Analytics for E-Learning in Higher Education., Flinders University, College of Science and Engineering
Terms of Use: This electronic version is (or will be) made publicly available by Flinders University in accordance with its open access policy for student theses. Copyright in this thesis remains with the author. You may use this material for uses permitted under the Copyright Act 1968. If you are the owner of any included third party copyright material and/or you believe that any material has been made available without permission of the copyright owner please contact copyright@flinders.edu.au with the details.
This study delves into the significant potential of Learning Management System (LMS) usage to influence student performance across diverse range of academic disciplines, utilising predictive data analytics to deepen our understanding and enhance educational outcomes. As E-Learning becomes increasingly central to educational delivery, understanding the complex effects of varied LMS interactions on student success is critical. Traditional analytics methods, focusing on customising learning materials based on student interactions within LMSs, often overlook the differential impacts of E-Learning across disciplines where engagement and effectiveness can vary significantly.
Structured around two pivotal questions, the research seeks to uncover:
RQ1: "How does Learning Management System (LMS) use across disciplines impact student performance?" This is explored through inquiries into how LMS usage varies across disciplines and its relation to student performance metrics, identification of specific LMS features as significant academic performance predictors, and the role of predictive data analytics models in pinpointing at-risk students across colleges. The question of dimensionality reduction's necessity in capturing essential LMS use aspects and its impact on predictive model performance is also addressed.
RQ2: "Do colleges differ significantly in approach and consistency?" This question investigates the variability of student individual differences, such as learning styles and engagement patterns, and their academic repercussions. It explores the implications of student behaviour variations captured through LMS data, for instructional design and student support services, and examines distinctive pedagogical approaches as evidenced by LMS data in relation to student engagement and performance.
Employing a broad suite of machine learning algorithms, including tree-based classifiers, probabilistic models, ensemble methods, and hybrid models, the study offers an in-depth analysis of discipline-specific engagement patterns and material consumption within the LMS. This methodological innovation facilitates a more detailed examination of how E-Learning implementations impact student performance across disciplines.
Findings reveal that distinct student profiles can be accurately identified for each college, enabling the prediction of course enrolment and the tailoring of machine learning approaches to predict performance within specific domains. Theoretically, the research advances a predictive data analytical model for E-Learning across disciplines, integrating insights from machine learning and E-Learning research to not only accurately predict student college affiliation through LMS data but also to leverage critical topic structure features for highlighting effective pedagogical strategies across disciplines.
Practically, this study equips educators with the insights needed to select and implement suitable E-Learning approaches, optimising teaching methods, material selection, and topic construction to meet disciplinary needs. Incorporating the instructional design methodology knows as ADDIE (Analysis, Design, Development, Implementation, and Evaluation) model, it provides a structured yet adaptable framework for designing and implementing E-Learning experiences, moving beyond generic solutions to offer tailored methodologies suited to distinct educational domains.
Moreover, by identifying effective algorithms for E-Learning tasks and elucidating effective pedagogical strategies in each discipline, the study enhances our understanding of E-Learning dynamics. It significantly advances predictive data analytics in E-Learning, employing innovative machine learning techniques and integrating established instructional design models to deliver actionable insights for educators and researchers, thus improving student performance across diverse educational settings.
Keywords: e-learning, predictive data analytics, machine learning, LMS design, discipline specific,
Subject: Computer Science thesis
Thesis type: Doctor of Philosophy
Completed: 2025
School: College of Science and Engineering
Supervisor: Giselle Rampersad