International Journal of
Information and Education Technology

Abstract

This study presents a predictive decision support system designed to assist Isabela State University in student admission decision-making through the application of data mining and deep learning techniques. The system was developed using a dataset of 24,278 anonymized student application records collected from seven university campuses, incorporating academic performance indicators, entrance examination scores, interview results, and selected demographic attributes. After comprehensive data preprocessing, including imputation, feature engineering, one-hot encoding, and normalization, a Multilayer Perceptron (MLP) classifier was trained to predict student admission outcomes. To contextualize model performance, a Logistic Regression classifier was implemented as a baseline using the same preprocessing pipeline and stratified 80:20 train-test split. Experimental results show that both models achieved comparable overall accuracy (83.16%), with Logistic Regression demonstrating higher precision and slightly higher Receiver Operating Characteristic-Area Under the Curve (ROC-AUC), while the MLP achieved higher recall and F1-score, indicating a more balanced identification of qualified applicants. Model evaluation was conducted using accuracy, precision, recall, F1-score, ROC-AUC, and confusion matrix analysis. In addition, a simple empirical fairness analysis was performed across available sensitive attributes, including gender, campus, and senior high school track. The results indicate no substantial performance disparities across major groups, suggesting reasonable empirical fairness within the evaluated dataset. Overall, the findings suggest that the MLP-based system can serve as a supportive, data-driven tool for university admission processes by providing consistent predictive insights while maintaining transparency and fairness considerations.