Building Effective Machine Learning Pipelines for Student Retention

Session Abstract: Student retention remains one of the most pressing challenges in higher education, and predictive analytics can provide institutions with the insights needed to take timely, impactful action. This session will share key lessons learned in building machine learning pipelines to identify at-risk students. Topics include: designing sustainable pipelines, applying dynamic feature selection, choosing and evaluating models, addressing class imbalance, interpreting results through probability ranking and performance metrics. We will also cover the role of automation and experiment tracking in improving efficiency and reproducibility. Rather than a step-by-step technical build, the session focuses on practical strategies, considerations, and pitfalls to avoid. Attendees will have a clear sense of what matters most when developing ML approaches for student success analytics.

Keywords: Student Retention, Machine Learning, Predictive Analytics, Model Interpretability