Predicting Dropout for Nontraditional Undergraduate Students: A Machine Learning Approach. (February 2023)

• Record Type: Journal Article
• Authors: Huo, Huade; Cui, Jiashan; Hein, Sarah; Padgett, Zoe; Ossolinski, Mark; Raim, Ruth; Zhang, Jijun
• Is Part Of: Journal of college student retention. Volume 24:Number 4(2023)
• Page Start: 1054
• Page End: 1077
• Publication Date: 2023-02
• Subjects: nontraditional students -- dropout rates -- machine learning -- decision trees; College dropouts -- Periodicals; College dropouts -- Prevention -- Periodicals; College attendance -- Periodicals; 378.16913
• DOI: 10.1177/1521025120963821
• Languages: English
• ISSNs: 1521-0251

Abstract

Student attrition represents one of the greatest challenges facing U.S. postsecondary institutions. Approximately 40 percent of students seeking a bachelor's degree do not graduate within 6 years; among nontraditional students, who make up half of the undergraduate population, dropout rates are even higher. In this study, we developed a machine learning classifier using the XGBoost model and data from the National Center for Education Statistics (NCES) Beginning Postsecondary Students (BPS) Longitudinal Study: 2012/14 to predict nontraditional student dropout. In comparison with baseline models, the XGBoost model and logistic regression model with features identified by the XGBoost model displayed superior performance in predicting dropout. The predictive ability of the model and the features it identified as being most important in predicting nontraditional student dropout can inform discussion among educators seeking ways to identify and support at-risk students early in their postsecondary careers.