Arko Chakrabartiroy
Class of 2026New York, New York
About
Projects
- "What factors predict the risk of dropout from college? Which Machine Learning model best predicts the risk of dropout?" with mentor Morteza (Working project)
Project Portfolio
What factors predict the risk of dropout from college? Which Machine Learning model best predicts the risk of dropout?
Started Mar. 6, 2025
Abstract or project description
A college degree is key in today’s economy for ensuring a successful career and life. College graduates earn more than high school graduates and college dropouts and also have considerably lower unemployment rates. Though college enrollment has been increasing across the world in recent years, college completion rates have been lagging behind – many college students do not persist in college beyond the first one or two years, and never graduate, despite often earning a significant number of credits. Hence, it is critical to understand the factors that contribute to dropping out of college, as only by knowing these factors, can we start to address the significant college dropout problem of today.
In my research paper, I utilize recent developments in machine learning models to explore the factors which are significantly associated with college persistence and completion. Employing a large and comprehensive dataset containing individual-level data on students enrolled in undergraduate degrees at an anonymous college and a variety of machine learning models, I isolate the variables which are significantly correlated with dropping out of college. This rich dataset incorporates demographic background, data on economic conditions like unemployment rate and inflation rate, and granular data on prior academic achievement such as courses enrolled and passed as well as grades earned in each course. I use three-quarters of the data to train the various machine learning models, while I use the remaining quarter of the data to test the models and find the most efficient model for college dropout.
The findings will throw extensive light on the factors which rank most highly in paving the road to student success in higher education. My expansive but rigorous analysis should be highly instructive for a broad range of audience – including academic researchers, public policy makers, university administrators and college students themselves.