THE STRATEGIC USE OF CONTEMPORANEOUS DATA AND PREDICTIVE MODELS FOR A CONTEXT-DEPENDENT LIFE-CYCLE APPROACH TO STUDENT RETENTION

Abstract

The impending enrollment crunch, stemming from a reduced college-age population, compounds the need for higher educational institutions to focus more attention on efforts to retain students throughout their undergraduate careers. In this paper, we present a framework for the strategic use of contemporaneous data and predictive models to identify first-time and transfer students at risk of withdrawal in their initial two years at a four-year, land-grant university across a diverse, multi-campus system. When moving toward a more context-dependent life-cycle approach to student retention, considerations must be given for the unique context each campus creates for its students and for the salient factors affecting student outcomes at each stage of their undergraduate careers. To this end, we highlight the practical decisions that have been made along the way in establishing this framework, from those regarding how datasets are created for campuses over time to those regarding the selection of predictive models at different timepoints during students’ tenure.

Figure 1. Confusion matrix by date of XGBoost predictive Student Success Model (SSM)*

* Note: Inconsistent population size due to 1,505 additional transfer students added to the predictive model on October 19, 2022 (for a total N of 9,999 first-time and transfer students).