Larasati, Imaning Dyah (2019) Prediksi Kelulusan Mahasiswa Berdasarkan Kinerja Akademik Menggunakan Metode Modified K-Nearest Neighbor (MK-NN). Sarjana thesis, Universitas Brawijaya.
Abstract
Perguruan tinggi merupakan salah satu penyelenggara pendidikan akademik bagi mahasiswa. Faktor penentu kualitas perguruan tinggi dapat dilihat melalui persentase kemampuan mahasiswa menyelesaikan studi dengan tepat waktu. Pada program studi Teknik Informatika Fakultas Ilmu Komputer (FILKOM) Universitas Brawijaya, terdapat permasalahan mengenai kinerja akademik mahasiswa dalam hal masa studi. Akademik FILKOM memiliki database yang menyimpan data akademik mahasiswa. Sehingga, dapat dilakukan data mining, salah satunya dengan mengklasifikasikan mahasiswa yang lulus tepat waktu dan tidak tepat waktu. Hasil klasifikasi digunakan untuk memprediksi kelulusan mahasiswa pada semester 5. Salah satu metode yang bagus dalam prediksi kelulusan adalah K-Nearest Neighbor (K-NN). Namun, terdapat metode yang memiliki akurasi lebih baik dari K-NN, yaitu Modified K-Nearest Neighbor (MK-NN). Sehingga, penelitian ini melakukan prediksi kelulusan mahasiswa berdasarkan kinerja akademik menggunakan metode MK-NN untuk mengetahui akurasi dengan melihat pengaruh dari beberapa parameter seperti nilai k, jumlah data latih dan komposisi data latih. Kemudian dilakukan perbandingan metode MK-NN dengan K-NN. Dari pengujian pengaruh nilai k didapatkan akurasi tertinggi sebesar 82% saat k=5. Kemudian akurasi terbaik yang didapatkan dari pengujian pengaruh jumlah data latih dan komposisi data latih masing-masing sebesar 85,25% dan 84%. Pada perbandingan metode MK-NN dengan K-NN didapatkan kesimpulan bahwa metode MK-NN memiliki akurasi lebih tinggi dibandingkan K-NN.
English Abstract
University is one of the academic education institutions. One important factor to determine the quality of higher education can be seen through the percentage of students' ability to complete studies on time. In the Brawijaya University Faculty of Computer Science (FILKOM) Informatics Engineering study program, there is problem about academic performance of students in term of the study period. Academic department of FILKOM has databases that store student academic data. Therefore, data mining can be done by classifying students who graduate on time and not on time. The classification results can be used to predict student graduation in a semester 5. One method that provides good accuracy in predicting graduation is K-Nearest Neighbor (K-NN). However, in other cases there is a better method than K-NN, that is Modified K-Nearest Neighbor (MKNN). So, this study predicts student graduation based on academic performance using MK-NN method to determine the accuracy by observing the effects of several parameters such as k value, amount of training data and composition of training data. Then compare the MK-NN method with the K-NN method. From the results of testing the effect of k values, the highest accuracy is 82% when k = 5. Then the best accuracy obtained from testing the effect of the amount of training data and testing the effect of the composition of training data is 85.25% and 84% respectively. The results of the comparison of the MK-NN method with the K-NN method, it was concluded that the MK-NN method had a higher accuracy than the K-NN method.
| Item Type: | Thesis (Sarjana) |
|---|---|
| Identification Number: | SKR/FILKOM/2019/103/051902273 |
| Uncontrolled Keywords: | prediksi, kelulusan, data mining, klasifikasi, K-Nearest Neighbor, Modified K-Nearest Neighbor |
| Subjects: | 300 Social sciences > 371 Schools and their activities; special education > 371.2 School administration; administration of student academic activities > 371.24 Schedules and school day > 371.244 School day > 371.291 2 School completion |
| Divisions: | Fakultas Ilmu Komputer > Teknik Informatika |
| Depositing User: | soegeng sugeng |
| Date Deposited: | 12 Jun 2020 07:24 |
| Last Modified: | 19 Oct 2021 09:07 |
| URI: | http://repository.ub.ac.id/id/eprint/168884 |
Preview |
Text
Imaning Dyah Larasati.pdf Download (3MB) | Preview |
Actions (login required)
 |
View Item |