Machine Learning for Fair and Accurate University Admission Prediction: A Case Study from the UAE

University admissions are high-stakes decisions that shape academic and professional futures. This study develops a machine learning framework to predict university admission outcomes using official Ministry of Education records and a public benchmark dataset. We evaluate logistic regression, random forest, gradient boosting, and random forest with SMOTE balancing, achieving up to 99.8% accuracy and minority-class recall above 94%. Beyond predictive performance, the study integrates fairness audits and explainability methods (feature importance, permutation importance, SHAP) to ensure transparency and accountability. Results reveal subgroup disparities across gender and nationality, underscoring the need for fairness-aware deployment. This work provides a methodological benchmark for researchers and practical guidance for policymakers, demonstrating how accuracy, fairness, and interpretability can be combined in algorithmic admission systems.