<![CDATA[Timely graduation is one of the important indicators of academic success of universities. However, the student monitoring process still faces obstacles in identifying students who have the potential to not graduate on time. This study offers a solution in the form of applying machine learning to predict graduation status while comparing the most effective algorithms. The research uses 750 data of students of the Informatics Study Program batch 2017 – 2021 sourced from the Academic Information System. The research methods include data preprocessing, handling missing values and outliers, encoding, normalization, forming Decision Tree, Random Forest, and Naïve Bayes models, as well as evaluation using accuracy, precision, recall, F1-score, and AUC-ROC. The results of the study show that Random Forest provides the best performance with an accuracy of 91.33%, precision of 90.48%, recall of 93.42%, F1-score of 91.93%, and an AUC-ROC of 0.952. Decision Tree obtained an accuracy of 84.67%, while Naïve Bayes was 80.00%. The Friedman test resulted in a value of p = 0.015 which showed a significant difference in performance. Thus, Random Forest is the best algorithm for predicting student graduation in this research dataset and has the potential to support academic early warning systems.]]>
