A Generalizable Neural Network for Predicting Student Retention

This research revisits a neural network-based decision support system presented at the ASEE 2009 Northeast Section Conference at the University of Bridgeport.The decision support system identified students who are "at-risk" of not retaining to their second year of collegiate study.At that time, the positive preliminary results presented were based upon a small out-of-sample dataset.This research updates the results using the full 2008 freshman cohort at Fort Lewis College (N = 800).Overall, the system correctly predicted retention for approximately 70% of the freshmen.Additionally, over three fourths of the retainers were correctly identified.This research offers a confirmed generalizable framework for the development of an early alert system that will allow institutions to identify accurately students who are "at-risk" of not retaining to their second year of study.The eleven salient predictors including ethnicity and socioeconomic status also offer insight into how an institution might improve its retention efforts.This research finally proffers possible means (e.g.STEM support classes, student clubs, peer-mentoring, etc... ) to assist students identified as "at-risk" in taking steps to mitigate their risk for non-retention.