Machine Learning Methods for Course Enrollment Prediction

Accurately forecasting course enrollment rates in higher education is of great concern in order to minimize unnecessary administrative costs as well as burden to both students and faculty. This research aimed to first recreate course enrollment predictions based on a conditional probability analysis using student data from San Diego State University (SDSU) and then to improve upon those predictions by applying classification and regression trees (CART) and random forest. The authors incorporated student demographic and academic information into algorithms to ascertain their influence on improving course enrollment prediction accuracy. They used these strategies to predict enrollment in General Chemistry at SDSU, a course with historically varied and large enrollment numbers in the multiple hundreds per semester. The authors then determined which factors were the most influential to the General Chemistry enrollment number using a variable importance metric derived from tree-based algorithms.  

fan-j--smJuanjuan Fan, Ph.D., is a Professor of Statistics and Data Science in the Department of Mathematics and Statistics, and serves as a Faculty Advisor at the Analytic Studies & Institutional Research (ASIR), at San Diego State University. Her research interests include survival analysis, decision trees and random forests, and observational study data. Working with her students and collaborators, she has published many papers assessing student success studies and solving various problems in educational data mining. Professor Fan received her Ph.D. in Biostatistics from the University of Washington.

You must be an AACRAO member to view this article.

Please login in or learn more about joining AACRAO.