Machine Learning Applications for At-Risk Students Prediction in Online Learning Environments: A Systematic Review and Quality Assessment

• In Event: Friday Roundtable Session 4:55 pm
  In Roundtable Session: Methods of Support for Online Learners (Table 11)

Abstract

This systematic review aimed to identify the current state of research on using machine learning applications for predicting at-risk students (i.e. those at risk of failure or dropouts) in online learning. We adopted a quality assessment framework to appraise the studies based on necessity, reproducibility, robustness, generalizability, and interpretability. A total of 61 studies were included in the review. Findings revealed (a) a growing trend of using classical Machine Learning and Deep Learning models for predicting at-risk students and (b) a wide range in the quality appraisal results of the reviewed studies. These findings suggest a need for a common understanding of terminology and techniques, as well as the development of guidelines for consistent reporting and result comparability.

Authors

• Hui Shi, Florida State University

• Nuodi Zhang, Florida State University

• Hunhui Na, University of North Carolina - Charlotte

• Secil Caskurlu, Florida State University