PREDICTING COLLEGE DROPOUT LIKELIHOOD BASED ON HIGH SCHOOL AND COLLEGE DATA: A MACHINE LEARNING APPROACH

Authors

  • Hemant Jain Data Analytics Department, University of Tennessee Chattanooga, Chattanooga, TN

DOI:

https://doi.org/10.20319/icstr.2024.5758

Keywords:

Machine Learning, Student Success in College, Dropout Rate, Prediction

Abstract

College dropout rate is a significant problem, especially in the US higher education system. Among all undergraduate students, up to 40% drop out before completing their degree. This significantly impacts students and Universities alike financially and in wasted efforts. Previous research shows that there are early indicators of college success in the high school record such as grades, attendance, disciplinary incidents, and ACT/SAT scores. Additionally, there are factors in college experiences, especially in the first year of college which impact the likelihood of dropout. However, to the best of our knowledge, there is no comprehensive model that can accurately predict the likelihood of college dropout and provide an early warning either in high school and/or the first year of college. We were fortunate to get access to longitudinal ten-year data of high school graduates of public schools in a county in the United States and were able to follow a subset of students who went to a specific public University. Based on more than one hundred variables from high school and college records and students’ final status we trained various machine learning models to predict the likelihood of student dropout and identify factors that play a significant role. Based on this information a prototype decision support system was developed and evaluated.

Downloads

Published

2024-06-04

How to Cite

Jain, H. (2024). PREDICTING COLLEGE DROPOUT LIKELIHOOD BASED ON HIGH SCHOOL AND COLLEGE DATA: A MACHINE LEARNING APPROACH. MATTER: International Journal of Science and Technology, 57–58. https://doi.org/10.20319/icstr.2024.5758