Comparison of Models for Classification of Learning Achievement of Middle School Students in Indonesia in 2019 using the Support Vector Machine Algorithm, Conditional Inference Trees, and Random Forest
Algoritma Support Vector Machine, Conditional Inference Trees, dan Random Forest untuk Klasifikasi Capaian Belajar Siswa SMP di Indonesia Tahun 2019
Indonesian JHS students' learning achievement is still low. During 2015-2019, the average national exam score for Indonesian JHS has always decreased. In the last national examination, the average national exam score was 52.82 and was included in the bad category. This certainly needs to be a concern for local governments and the education office. Therefore, it is necessary to form a classification model that can be used to identify cities/districts in Indonesia which are categorized as bad or enough. This study discusses the comparison of models for the classification of learning achievement categories as seen from the average 2019 JHS results in 514 districts/cities in Indonesia using the Support Vector Machine (SVM), Conditional Inference Trees (Ctree), and Random Forest (RF) algorithms. The three algorithms were chosen because of their respective advantages, namely the SVM algorithm is known to be very powerful, Ctree as an improvement from the usual decision tree, and RF to represent ensemble learning. The independent variables used are education budget, classroom conditions, school accreditation, and teacher qualifications. From the results of this study, it has been found that the SVM algorithm produces the highest accuracy (0,80), recall (0,97), kappa statistics (0,38), and F1-score (0,87) compared to the Ctree and RF algorithms, while only precision (0,80) has the same value as the Ctree algorithm. So, the SVM algorithm produces the best model for the classification of district/city learning achievement categories in Indonesia based on education budget, classroom conditions, school accreditation, and teacher qualifications.
Bahasa Indonesia
Capaian belajar siswa yang dilihat melalui rata-rata nilai Ujian Nasional (UN) Sekolah Menengah Pertama (SMP) tahun 2019 menunjukkan bahwa masih rendahnya kualitas sistem pendidikan di Indonesia. Selama lima tahun terakhir (2015-2019), rata-rata nilai UN SMP Indonesia selalu mengalami penurunan. Pada UN terakhir, rata-rata nilai UN SMP Indonesia 2019 sebesar 52,82. Hal ini tentu perlu menjadi perhatian bagi pemerintah daerah dan dinas pendidikan. Oleh karena itu, perlu dibentuk sebuah model yang dapat mengurangi permasalahan tersbeut. Penelitian ini membahas pengklasifikasian kategori rata-rata nilai hasil UN SMP tahun 2019 di 514 kabupaten/kota di Indonesia menggunakan algoritma Support Vector Machine (SVM), Conditional Inference Trees (Ctree), dan Random Forest (RF). Ketiga algoritma itu dipilih karena kelebihannya masing-masing, yaitu algoritma SVM dikenal sangat powerful, Ctree sebagai perbaikan dari decision tree biasa, serta RF untuk mewakili ensemble learning. Variabel independen yang digunakan yaitu: (1) anggaran pendidikan, (2) kondisi ruang kelas, (3) akreditasi sekolah, dan (4) kualifikasi guru. Sedangkan variabel dependen yang digunakan adalah adalah rata-rata nilai hasil UN SMP dikategorikan menjadi dua, yaitu cukup dan kurang. Dari hasil penelitian ini telah didapatkan bahwa algoritma SVM mengasilkan accuracy (0,80), recall (0,97), kappa statistics (0,38), dan F1-score (0,87) yang tertinggi dibandingkan algoritma Ctree dan RF, sedangkan hanya precision (0,80) yang nilainya sama dengan algoritma Ctree. Maka, algoritma SVM adalah metode yang terbaik untuk mengklasifikasikan kategori rata-rata nilai hasil UN SMP.
Afiliasi berasal dari OpenAlex; metadata OAI-PMH jurnal tidak memuatnya. Penulis tanpa ID OpenAlex dicocokkan berdasarkan nama, sehingga dua orang bisa tergabung atau satu orang bisa terpisah.