Prediction of Student Graduation Using Decision Tree Method

Abstract

Students are one of the important parameters in the evaluation of study programs. Student attendance, student achievements, and graduation profiles must receive serious attention. More and more students are accepted every year, but not a few are able to complete their studies on time. Predictions for the current graduation rate are needed, considering the number of students experiencing delays is assessed to have reached 92%. Based on the data studied, there were 25 students in one class, only 2 students who graduated on time were confirmed to be late. To make this prediction, the writer uses Decision Tree. Testing student graduation using validation data as many as 110 lines of data, consisting of regular students A and regular B. There are 50 students graduating on time and 60 students graduating late. Then the technique of balancing the number of classes on the data is carried out, namely the Synthetic Minority Oversampling Technique (SMOTE), so that the number of validation data increases to 120 rows of data. This test resulted in a score of 96.67%, 96.67% precision, 96.67% recall, and 96.67% f1-score. Based on the ROC curve, the evaluation results that have been achieved in predicting future student graduation using the Decision Tree method, including a very good classification.