Application of Students Graduation Prediction Model Using Decision Tree C4.5 Algorithm and Synthetic Minority Oversampling Technique (SMOTE)

An important achievement for students and higher education institutions is timely graduation. Unfortunately, a large number of students fail to graduate on time, which can have a negative impact on the accreditation of higher education institutions. Academic authorities can improve the timely graduation rate by developing policies and regulations with the help of student graduation prediction. Using academic data, data mining is an efficient way to predict graduation. Consequently, the Decision Tree C4.5 approach is used in this study to predict student graduation. This study also uses the SMOTE oversampling technique to address class imbalance in minority data and the mean method to manage missing variables. According to the findings, the first-semester grade point average (GPA) has the greatest impact on graduation. Overall GPA, GPA from the fourth semester, and GPA from the second semester are additional significant factors. An accuracy of ${8 4. 4 \%}$, precision of ${8 8. 8 \%}$, recall of ${8 3. 4 \%}$, and F1 score of 0.852 with N value of $100 \%$ for SMOTE were obtained through model testing using 5-Fold Cross Validation with SMOTE. The results of the model without SMOTE are $81.6\%$ accuracy, $87\%$ precision, ${8 4. 8 \%}$ recall, and an ${F 1}$ score of ${0. 8 5 0}$. Furthermore, with ${8 2. 8 \%}$ accuracy, the highest parameter values for tree depth are 5 and 10.