<![CDATA[This study aims to predict student graduation outcomes by utilizing demographic and academic variables from the Student Performance Dataset. The analysis was conducted using the Logistic regression method, selected for its ability to handle binary outcomes and provide clear interpretability of predictor contributions. The research process included data preprocessing, removal of variables G1 and G2 to prevent data leakage, and conversion of the final grade (G3) into a binary graduation label. The model was evaluated using accuracy, logistic loss, and a confusion matrix to measure predictive reliability and classification stability. The results indicate that the model achieved an accuracy of 78.85% with a logistic loss value of 0.412, demonstrating stable performance and good generalizability. These findings suggest that simple demographic and academic attributes—such as age, study time, prior failures, and attendance—play a significant role in predicting graduation likelihood. Overall, the study confirms that Logistic regression is an effective approach for educational data analysis and can be utilized by schools to identify at-risk students and design more targeted instructional interventions.]]>
Copyrights © 2025
