Prediksi Kelulusan Mahasiswa Berdasarkan Kinerja Akademik Menggunakan Metode Modified K-Nearest Neighbor (MK-NN)

Imaning Dyah Larasati (Fakultas Ilmu Komputer, Universitas Brawijaya)
Ahmad Afif Supianto (Fakultas Ilmu Komputer, Universitas Brawijaya)
Muhammad Tanzil Furqon (Fakultas Ilmu Komputer, Universitas Brawijaya)

Jurnal Pengembangan Teknologi Informasi dan Ilmu Komputer
Vol 3 No 5 (2019): Mei 2019

In the Brawijaya University FILKOM Informatics Engineering study program, the academic performance of students in terms of study period is still a problem. In FILKOM's academic database, there are student academic data. The data can be carried out data mining by predicting students' graduation in the 5th semester. K-NN is a good method for predicting graduation. However, there is a method that has better accuracy than K-NN has been found in other cases, that is MK-NN. Therefore, this study using M-NN method for predict students' graduation based on academic performance by testing includes testing the effect of k value, the number of training data and the composition of training data. Furthermore, comparing the accuracy produced by MK-NN and K-NN methods. The highest accuracy of testing the effect of the value of k is when k = 5, which is equal to 82%. The highest accuracy from testing the effect of number of training data and the composition of training data reached 85,25% and 84%. From the comparison of the accuracy of MK-NN and K-NN it was concluded that MK-NN produced better accuracy than K-NN.