A Cause-Aware Graduation Outcome Prediction Framework for Academic Decision Support
Year: 2026
Type: conference-paper
Abstract: Timely graduation is a critical indicator of educational quality, yet delayed graduation remains a persistent challenge in higher education. Although machine learning has been widely applied to graduation outcome prediction, most existing studies rely on coarsegrained formulations that treat delayed graduation as a homogeneous outcome, providing limited actionable insight for academic intervention. This paper proposes a cause-aware graduation outcome prediction framework that explicitly distinguishes heterogeneous causes of delayed completion. A five-class graduation outcome taxonomy is introduced, capturing graduation quality for on-time graduates and separating language certificationrelated delays from coursework-related delays. Based on this formulation, an XGBoost-based multi-class prediction model is developed to effectively handle heterogeneous educational data under class imbalance. To enhance transparency, the framework supports feature-level interpretation of model behavior. Experiments conducted on academic records of 787 undergraduate students demonstrate strong and stable predictive performance. The results show that causeaware outcome modeling significantly improves the interpretability and practical value of graduation outcome prediction for academic decision support.
Institution: Industrial University of Ho Chi Minh City
Language: English
Subfield: Computer Science Applications
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7166136241
