Comparison of SMOTE Sampling Based Algorithm on Imbalanced Data for Classification of New Student Admissions

Abstract

One of the efforts to get quality students is through selection. The selection process must be balanced with a strategy so that the selected students are truly qualified. Classification techniques can be used to see the history of new student admissions who are accepted with the student’s lecture history. There are many classification algorithms that can be used, so comparisons need to be made to see the best performance of the algorithm. The classification algorithm used is Decision Tree C4.5, K-Nearest Neighbor, Naïve Bayes and Neural Network. The data used are 546 records in the imbalanced data category. So we need the Smote algorithm to make the data balanced so as not to result in misclassification. The classification results were tested using the Confusion Matrix, ROC and Geometric Mean (G-Mean) as well as a T-Test. The comparison results show that the best performance is on the K-Nearest Neighbor algorithm with an accuracy value of 84.99%, AUC of 0.700, G-Mean 62.95% and the T-test produces a significant different from other algorithms.