Citation

Manzoor, Shahida Raihan (2024) Learning style prediction for online students using Support Vector Machine approach. Masters thesis, Multimedia University.

Full text not available from this repository.

Abstract

The COVID-19 pandemic significantly reshaped education, accelerating the global shift to online learning or e-learning. As institutions adopted virtual platforms, factors such as technological infrastructure, instructor engagement, learning styles (LS), personality traits, motivation, and social interaction emerged as key determinants of online learning effectiveness. Prior research highlights that aligning online instruction with students’ LS and personality traits enhances academic performance, engagement, and satisfaction but studies on LS prediction have not included personality traits in their models despite the impact of personality on the overall e-learning experience. While some Learning Management Systems (LMSs) support LS-based personalization, none currently implement automated LS prediction frameworks that consider personality traits. This study addresses that gap by proposing a machine learning (ML)-based model to predict students’ VARK learning styles using personality traits, performance, and behavioral data. Focusing on 78 students from Multimedia University’s FCI department, the research employed K-means clustering and linear regression during preprocessing, confirming the absence of natural clusters and justifying the use of supervised binary classification. Multiple ML models were evaluated, including Naïve Bayes, Decision Trees, Random Forests, and Support Vector Machines (SVM). The unoptimized SVM outperformed others in 5 of 7 evaluation metrics—accuracy, AUC, classification error, precision, and F1 score—indicating its strong baseline performance. Subsequent optimization using Principal Component Analysis (PCA) and an SVM with a Radial Basis Function (RBF) kernel (C=1, gamma=0.1) achieved the highest accuracy of 78%. The final model or the eVLS model, successfully classifies learners as Kinesthetic or Non-Kinesthetic with 78% accuracy, based on Five-Factor personality traits, behavioral, and performance data. This study also discusses the possible integration of the eVLS model into LMS systems and its potential to yield a multitude of favourable outcomes within the academic realm. This sort of integration has the potential to offer personalized learning, reduce learning barriers, and improve academic outcomes, satisfaction, and learner engagement.

Item Type:	Thesis (Masters)
Additional Information:	Call No.: Q325.5 .S53 2024
Uncontrolled Keywords:	Machine learning
Subjects:	Q Science > Q Science (General) > Q300-390 Cybernetics
Divisions:	Faculty of Computing and Informatics (FCI)
Depositing User:	Ms Nurul Iqtiani Ahmad
Date Deposited:	19 Jan 2026 04:20
Last Modified:	19 Jan 2026 04:20
URII:	http://shdl.mmu.edu.my/id/eprint/15190

Downloads

Edit (login required)