Retention is Everyone's Job: Decision Tree Models as a Method to Engage Retention Stakeholders on Campus

Abstract

Regression-based retention models produce excellent predicted student outcomes to alert campus faculty and staff to potential retention risks. However, models using decision tree classification allow stakeholders to see how retention breaks down for categories of students in their domains. We implemented logistic regression modeling for freshmen over four years and saw success in identifying students for interventions, but quickly after implementing decision tree classification models, we were able to engage retention stakeholders (admissions, coaches, academic units) and gather qualitative data on how to best tailor interventions in future years. A complementary approach of using both types of models is recommended.