Researchers from CCRC have previously recommended a total of 13 first-year academic-momentum metrics as leading Key Performance Indicators (KPIs) for long-term outcomes at community college. This study examines the predictive power of these metrics for degree completion using seven different supervised Machine Learning (ML) algorithms and transcript-level data obtained from 36 institutions in two states. The presentation described findings that indicate the KPIs explain roughly 75-80 percent of student’s likelihood of degree completion, and the predictive performance remained stable across different demographic groups. These findings suggest that the KPIs are informative near-term outcomes that predict degree completion. This study also found that a traditional logistic regression model performed as well as commonly used ML models, suggesting that a marginal gain in predictive performance by ML is small when a model uses transcript-level student data.
Assessing the Completeness of Academic Momentum Theory for Community College Degree Completion using Supervised Machine Learning Techniques
Date and Time:
March 21, 2019 11:30 a.m.–1 p.m.
Location:
Kansas City, MO
Venue:
Kansas City Marriott Downtown
Associated Papers
Participants
Takeshi Yanagiura
Senior Research Assistant
Community College Research Center
Community College Research Center