What's Happening?
A study has been conducted to evaluate ensemble models for predicting student performance in higher education using multimodal data. The research focuses on fully online Information Technology programs, where attrition poses a threat to institutional sustainability and educational equity. The study integrates Moodle interactions, partial grades, demographic data, SMOTE balancing techniques, and stacking ensemble models to predict student performance. The framework is evaluated using a cohort of 2,225 students enrolled in online IT programs at the Universidad Estatal de Milagro. The study aims to determine whether stacking models outperform individual models like SVM and Random Forest and analyze fairness before and after balancing.
Why It's Important?
The study is significant as it addresses the challenge of early prediction of student performance, which is crucial for enhancing academic outcomes and reducing attrition rates in online education. By integrating various data sources and advanced modeling techniques, the research provides valuable insights into student behavior and learning patterns. This approach can lead to more equitable and personalized intervention strategies, promoting student retention and achievement. The study also highlights the importance of fairness and interpretability in educational predictive modeling, ensuring that predictions are not only accurate but also equitable and understandable.
What's Next?
The next steps involve translating the predictions into effective personalized interventions that address each learner's unique needs. Institutions can use data-driven insights to proactively respond to academic challenges and support student retention. The study's findings may lead to the development of more robust and nuanced predictive models, contributing to a more equitable and personalized early prediction framework in higher education.
