<![CDATA[Student dropout poses a critical challenge for higher education institutions, impacting academic quality and institutional evaluation. This study aims to develop an early warning system to detect student dropout risks at the Faculty of Tarbiyah, UIN Sjech M. Djamil Djambek Bukittinggi. Employing a quantitative research and development (R&D) approach with a Waterfall framework, the system utilizes an Artificial Neural Network (ANN) algorithm, specifically a Multilayer Perceptron (MLP) architecture. The model was trained on 596 student academic records, evaluating five primary independent variables: cohort year, GPA, total credit hours taken, total credit hours passed, and number of repeated courses. The evaluation using a confusion matrix demonstrates that the ANN model achieves a high weighted average accuracy of 94%. Furthermore, the model is successfully integrated into a stateless web platform using the Flask framework, enabling instantaneous prediction without permanent database storage to rigidly uphold student data privacy. This system provides a significant practical contribution by assisting academic advisors in initiating timely, data-driven interventions to prevent student attrition securely. ]]>
Copyrights © 2026
