Predicting College Student Dropouts with Machine Learning

Abstract

In the U.S., 33% of college students do not graduate, with 24% dropping out in their first year [2]. Various factors contribute to these high dropout rates, with finances as a significant one: some sources estimate that as high as 51% of college students drop out because of a lack of funds, a statistic that is confirmed by the machine-learning models used in this experiment [6]. In this study the neural network and random forest classifier machine learning models are used to predict whether a student will graduate or drop out. The “Predict students dropout, academic success” dataset was used [4]. This dataset provides various pieces of information about students, including gender, age of enrollment, and dropout status. The random forest classifier had an accuracy of 92.05% and the neural network had an accuracy of 91.71%, with academic and finance related factors contributing most to the dropout rate predictions. The ability to predict college student dropout status using machine learning is crucial, as it allows colleges to recognize students at risk of dropping out, enabling institutions to provide additional resources.