Early Student Dropout Prediction in Online Learning Using Explainable Machine Learning and Deep Learning Models

Annamalai, Shakti Siva Raman (2025) Early Student Dropout Prediction in Online Learning Using Explainable Machine Learning and Deep Learning Models. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (2MB) | Preview

Abstract

Online learning is now part and parcel of contemporary learning, but it is facing significantly greater rates of dropouts than in the real classrooms. In this study, an early prediction system that has been developed is aimed at identifying potentially at-risk students at the start of the first few weeks of a module so that a tutor can intervene before the situation gets out of control. Based on the Open University Learning Analytics Dataset (OULAD) of demographic attributes and the early engagement behavior, a number of predictive models were applied, as follows, Logistic Regression, Decision Tree, XGBoost, and an Artificial Neural Network. Out of them XGBoost was the most successful model on the whole, with good predictive stability and can be used as an early warning tool. SHAP explainability tools were utilized to achieve greater transparency and interpretability and determine the factors that contribute to dropout the most, with the early click activity and previous academic achievements seeming to be the most important factors. A probability-based risk stratification mechanism was also included in the system that allowed grouping students into valuable risk groups to facilitate targeted intervention. In general, the study represents a feasible, understandable, and scalable early-warning model that can be implemented in learning institutions to enhance student retention and the level of learner support. The work can be continued in the future to expand the feature set and investigate sequential modelling strategies to gain further insights into behavior.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Staikopoulos, Athanasios UNSPECIFIED
Uncontrolled Keywords:	Early Dropout Prediction; Machine Learning; Deep Learning; XGBoost; Logistic Regression; Artificial Neural Network (ANN) SHAP Explainability; Student Engagement; Predictive Modelling; Risk Stratification
Subjects:	L Education > L Education (General) L Education > LB Theory and practice of education > LB2300 Higher Education L Education > LC Special aspects / Types of education > E-Learning Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	07 Sep 2026 08:47
Last Modified:	07 Sep 2026 08:47
URI:	https://norma.ncirl.ie/id/eprint/9845

Actions (login required)

View Item