Perbandingan Algoritma Regresi Logistic Dan Neural Network Pada Prediksi Nilai Hasil Pembinaan Dan Kelulusan Tepat Waktu

Abstract

INTISARI

Penelitian ini didasari pada keinginan memanfaatkan informasi akademis mahasiswa yang tinggal di asrama yang memiliki pendidikan karakter dengan program pembelajaraan milik Universitas Muhammadiyah Yogyakarta yang disediakan untuk sebagian mahasiswanya. Hubungan antara pembinaan di asrama mahasiswa dengan prestasi di kampus belum pernah diteliti secara khusus. Penelitian sebelumnya yang penulis temukan menjelasakan hubungan antara nilai di kampus dan kelulusannya. Adanya visi asrama yang salah satunya adalah prestasi studi juga tersedianya data Nilai pendaftaran hingga raport hasil pembelajaran di Asrama serta data kelulusan di kampus, sehingga penulis ingin melihat apakah mahasiswa asrama dapat lulus tepat waktu di kampus, dibutuhkan data mining untuk memprediksi, dipilihlah algoritma Regresi Logisitic dan Neural Network. Dari hasil pengolahan data angkatan tahun 2014-2015 yang digunakan untuk training dan testing, didapatkan hasil dari 5x iterasi k-fold cross validation untuk Regresi Logistic dengan akurasi 65 % dan Neural Network 69%. Dengan begitu algoritma Neural network cendrung lebih baik Regresi Logistic.

 Kata kunci — data mining, kelulusan, klasifikasi, neural network, prediksi, regresi logistic

ABSTRACT

This research is based on the desire to utilize the academic information of students living in dormitories who have character education with the learning program of the University of Muhammadiyah Yogyakarta provided for some of its students. The relationship between development in student dormitories with achievements on campus has not been specifically examined. Previous research that the authors found explained the relationship between grades on campus and graduation. The existence of a dormitory vision, one of which is the achievement of the study as well as the availability of data Registration value to report cards of learning outcomes at the Dormitory as well as graduation data on campus, so the writer wants to see whether boarding students can graduate on time on campus, data mining is needed to predict, chosen Logistic Regression algorithm and Neural Network. From the results of the 2014-2015 batch data processing used for training and testing, the results of 5 times the k-fold cross validation iteration for Logistic Regression with an accuracy of 65% and a 69% Neural Network. Thus the Neural network algorithm tends to be better than Logistic Regression.

 Keywords —  data mining, graduation, klasification, neural nework, prediction, regresi logistic.