Machine Learning for Graduation Prediction in Higher Education: A Systematic Review with a Bio-Inspired Optimization...

Area of Science:

• Educational Data Mining
• Machine Learning in Higher Education
• Predictive Analytics

Background:

• Timely graduation and degree completion are crucial metrics for student success and institutional performance.
• Machine learning (ML) offers potential for predicting student graduation outcomes.
• A systematic review is needed to assess current ML applications in graduation prediction.

Purpose of the Study:

• To systematically review machine learning approaches for graduation prediction.
• To analyze predictive targets, pipeline components, scalability, and bio-inspired optimization in existing studies.
• To identify research gaps and opportunities for advancing educational analytics.

Main Methods:

• A PRISMA-based systematic literature review was conducted.
• Searches were performed in Web of Science Core Collection and Scopus.
• 25 studies published between 2021 and 2025 were included after screening 278 records.

Main Results:

• Most studies treated graduation prediction as a supervised classification task, primarily using academic performance data.
• Tree-based and ensemble models were frequently employed, with common focus on feature selection, explainability, and hyperparameter optimization.
• Bio-inspired optimization (e.g., PSO, GA, ACO) was underutilized, and scalability issues were noted due to single-institution datasets and limited external validation.

Conclusions:

• There is a significant opportunity to develop Bio-Inspired Educational Analytics by focusing on scalable feature selection and efficient hyperparameter optimization.
• Future research should investigate lightweight, hybrid, and multi-objective metaheuristics for improved accuracy, interpretability, fairness, and transferability in graduation prediction.
• Advancing ML in higher education requires addressing limitations in scalability and exploring novel optimization strategies.