Leveraging Student Outcome Predictive Analytics to Mitigate Institutional Drop-out and Performance Risks in Higher Education Curriculum Projects

• Abiodun Okunola

  Ladoke Akintola University Technology

  Author

Student attrition and underperformance remain persistent challenges in higher education, with significant socioeconomic consequences for individuals and institutions. Despite growing interest in data-driven solutions, existing predictive models often suffer from limited generalizability, lack of interpretability, and failure to capture the temporal dynamics of student learning behavior. This study presents a comprehensive predictive analytics framework designed to identify at-risk students early and enable timely, targeted interventions within higher education curriculum projects. Using retrospective academic data from 8,267 undergraduate student records spanning multiple disciplines, we implemented and compared a suite of machine learning algorithms, including XGBoost, Random Forest, Support Vector Machines, and Neural Networks, alongside a stacked ensemble model incorporating performance-based weighting. The proposed framework achieved an overall accuracy of 89.4% in predicting student dropout risk by the fourth week of the semester, with a precision of 91.2% and recall of 87.6% for the at-risk category. Feature importance analysis identified prior academic achievement, first-year course performance, and LMS engagement metrics as the most significant predictors. The framework addresses critical research gaps by integrating explainable AI techniques through SHAP analysis to enhance model interpretability for institutional stakeholders. The findings demonstrate that predictive analytics, when operationalized through accessible and transparent decision-support systems, can substantially improve early warning capabilities and facilitate proactive student retention strategies in higher education.

• PDF

Copyright (c) 2026 Abiodun Okunola (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.