<p>Forecasting whether admitted candidates will ultimately enroll remains a critical challenge in university admissions, influencing recruitment strategies and institutional planning. This study proposes a data-driven modeling framework for enrollment prediction within a Vietnamese higher education institution, leveraging comprehensive administrative records from regular undergraduate programs. A dataset of 121,051 admission records was used to train and evaluate a suite of algorithms, including traditional classifiers such as logistic regression, Naive Bayes, decision trees, support vector machines, K-nearest neighbors, random forest, and gradient boosting, as well as deep learning architectures, namely multilayer perceptron and convolutional neural networks. These models were selected for their capacity to handle large-scale classification tasks involving complex feature interactions. Experimental results show that the multilayer perceptron model achieved the best performance, with 79.68% accuracy and an F1-score of 88.46% on the held-out 2023 test set. Permutation importance analysis revealed that program preference order, admission cutoff score, and high school graduation year were the most influential predictors of enrollment behavior. The findings contribute a scalable and interpretable solution for enrollment decision modeling and offer practical insights to support predictive analytics and strategic planning in autonomous university admission systems across diverse educational contexts.</p>