ANALISIS KOMPARATIF ALGORITMA KLASIFIKASI DATA MINING UNTUK PREDIKSI STATUS KELULUSAN MAHASISWA TEKNIK INFORMATIKA

Iqlimah Attyyatullatifah; Mia kamayani;

ABSTRAK

Analisis Komparatif Algoritma Klasifikasi Data Mining untuk Prediksi Status Kelulusan Mahasiswa Teknik Informatika

Iqlimah Attyyatullatifah

Penyelesaian studi tepat waktu merupakan indikator penting dalam menilai kualitas dan kompetensi
lulusan perguruan tinggi. Namun, tidak semua mahasiswa mampu menyelesaikan studi mereka
sesuai jadwal yang telah ditetapkan, sehingga diperlukan upaya untuk memprediksi potensi
keterlambatan kelulusan. Penelitian ini bertujuan untuk membandingkan kinerja empat algoritma
klasifikasi data mining, yaitu Decision Tree, Naïve Bayes, K-NN, dan SVM, dalam memprediksi
status kelulusan mahasiswa Teknik Informatika di Universitas Muhammadiyah Prof. Dr. Hamka.
Penelitian menggunakan dataset mahasiswa angkatan 2018–2020 dengan total 500 data yang
mencakup atribut akademik seperti IPS semester awal hingga IPK, serta penambahan atribut
demografis seperti jenis kelamin, asal sekolah, dan kota asal. Evaluasi dilakukan menggunakan
metode K-Fold Cross Validation (K=5), Confusion Matrix, dan Receiver Operating Characteristic
(ROC). Hasil penelitian menunjukkan bahwa algoritma K-NN memiliki performa terbaik dengan
akurasi 92%, presisi 92%, dan recall 90%, diikuti oleh Decision Tree dengan akurasi 90%, presisi
87%, dan recall 90%. SVM mencatatkan akurasi 84%, presisi 90%, dan recall 73%. Sementara itu,
Naïve Bayes memiliki akurasi 83%, presisi 80%, dan recall 83%. Hasil ini menunjukkan bahwa K-
NN unggul dalam mengenali pola data yang kompleks dibandingkan algoritma lainnya. Penelitian ini menyimpulkan bahwa penggunaan dataset yang lebih besar dan fitur yang relevan dapat
meningkatkan akurasi prediksi status kelulusan mahasiswa.

Kata kunci: prediksi kelulusan, perbandingan algoritma, klasifikasi, orange, data mining

Comparative Analysis of Data Mining Classification Algorithms for Predicting Graduation Status of Informatics Engineering Students

Iqlimah Attyyatullatifah

Timely completion of studies is an important indicator in assessing the quality and competence of
university graduates. However, not all students are able to complete their studies according to the
predetermined schedule, so efforts are needed to predict potential graduation delays. This study
aims to compare the performance of four data mining classification algorithms, namely Decision
Tree, Naïve Bayes, K-NN, and SVM, in predicting the graduation status of Informatics Engineering
students at Muhammadiyah Prof. Dr. Hamka University. The research uses a dataset of students
from 2018-2020 with a total of 500 data which includes academic attributes such as initial semester
IPS to GPA, as well as additional demographic attributes such as gender, school origin, and
hometown. Evaluation was carried out using the K-Fold Cross Validation (K=5), Confusion Matrix,
and Receiver Operating Characteristic (ROC) methods. The results showed that the K-NN algorithm
performed best with 92% accuracy, 92% precision, and 90% recall, followed by Decision Tree with
90% accuracy, 87% precision, and 90% recall. SVM recorded 84% accuracy, 90% precision, and
73% recall. Meanwhile, Naïve Bayes has 83% accuracy, 80% precision, and 83% recall. These
results show that K-NN excels in recognizing complex data patterns compared to other algorithms.
This study concludes that the use of larger datasets and relevant features can improve prediction
accuracy.

Keywords: graduation prediction, algorithm comparison, classification, orange, data mining