Predicting First-Year Retention in College Using Machine Learning and Traditional Models

• In Event: Data-Driven Insights into Student Retention, Attainment, and Beyond

Abstract

This study examines the application of several machine learning and logistic regression models to predict first-year student retention at a large urban university. It addresses two research questions: (1) What are the key student factors influencing first-year retention? and (2) What is the most effective predictive model for identifying students at risk of not being retained? Preliminary findings suggest that combining machine learning and logistic regression enhances predictive accuracy. The study is significant in several ways: it addresses class imbalances in the dependent variable, it engages the discourse between logistic regression and machine learning as an openly held tension, and it offers both theoretical and actionable models based on its findings to improve student success. 

Authors

• Mario Martinez, University of Houston

• Sanga Kim, University of Houston

• John Emmons, University of Houston

• Caroline Neary, University of Houston