PREDICTING COURSE ENROLLMENT WITH MACHINE LEARNING AND NEURAL NETWORKS: A COMPARATIVE STUDY OF ALGORITHMS.

The digitization and collection of big data by higher education institutions can help administrators make informed decisions about resource allocation, specifically in enrollment management. This study explores the use of machine learning and neural network algorithms to predict future student enrollments in courses. Real data from the Arab American University in Palestine (AAUP) was used. Eight machine learning algorithms, in addition to the Multilayer Perceptron (MLP) neural network model, were used to predict which students are most likely to enroll in a specific course. Results show that ensemble-based and bagging algorithms outperform other classifiers, including neural network models, for individual-level prediction of student enrollments. The Random Forest algorithm achieves the highest accuracy of 94% and an F1 score of 79% after applying under-sampling techniques on the highly imbalanced dataset. The study recommends future research to develop a generalized model for predicting enrollment in any course at AAUP and highlights the effectiveness of these techniques for improving resource allocation and student support.