IMPLEMENTATION OF K-NEAREST NEIGHBOR AND GINI INDEX METHOD IN CLASSIFICATION OF STUDENT PERFORMANCE

Penerapan Metode K-Nearest Neighbor Dan Gini Index Pada Klasifikasi Kinerja Siswa

Abstract

Predicting student academic performance is one of the important applications in data mining in education. However, existing work is not enough to identify which factors will affect student performance. Information on academic values ​​or progress on student learning is not enough to be a factor in predicting student performance and helps students and educators to make improvements in learning and teaching. K-Nearest Neighbor is a simple method for classifying student performance, but K-Nearest Neighbor has problems in terms of high feature dimensions. To solve this problem, we need a method of selecting the Gini Index feature in reducing the high feature dimensions. Several experiments were conducted to obtain an optimal architecture and produce accurate classifications. The results of 10 experiments with values ​​of k (1 to 10) in the student performance dataset with the K-Nearest Neighbor method showed the highest average accuracy of 74.068 while the K-Nearest Neighbor and Gini Index methods showed the highest average accuracy of 76.516. From the results of these tests it can be concluded that the Gini Index is able to overcome the problem of high feature dimensions in K-Nearest Neighbor, so the application of the K-Nearest Neighbor and Gini Index can improve the accuracy of student performance classification better than using the K-Nearest Neighbor method.