Model Klasifikasi Kesesuaian Artikel Pada Jurnal SINTA Berdasarkan Metadata Menggunakan Term Frequency–Inverse Document Frequency dan Naïve Bayes

Abstract

Pertumbuhan publikasi ilmiah di Indonesia menimbulkan kebutuhan akan metode yang efisien untuk membantu peneliti mengklasifikasikan dan menentukan jurnal yang sesuai bagi artikel akademik. Berbagai penelitian menunjukkan bahwa metode machine learning, khususnya Naïve Bayes, efektif dalam tugas klasifikasi teks berbahasa Indonesia. Namun, penelitian yang secara khusus memanfaatkan metadata artikel untuk menentukan kesesuaian artikel terhadap jurnal terindeks SINTA masih terbatas, khususnya terkait integrasi TF–IDF dan evaluasi berbasis cross-validation. Penelitian ini bertujuan mengembangkan model klasifikasi kesesuaian artikel pada jurnal SINTA berdasarkan metadata menggunakan Term Frequency–Inverse Document Frequency dan Naïve Bayes. Dataset terdiri atas 1.200 metadata artikel mencakup judul, abstrak, dan kata kunci, yang dikumpulkan melalui crawling manual terhadap jurnal-jurnal bidang teknologi pada portal SINTA. Tahapan penelitian meliputi pengumpulan data, prapemrosesan teks (case folding, translasi, tokenisasi, stopword removal, dan stemming), penggabungan metadata, ekstraksi fitur menggunakan TF–IDF, serta penerapan algoritma Naïve Bayes dengan skema 5-fold cross-validation. Evaluasi berdasarkan confusion matrix menunjukkan bahwa model mencapai accuracy 0,7058, precision 0,6977, recall 0,7133, dan F1-score 0,7065. Hasil ini menegaskan bahwa Naïve Bayes mampu memberikan performa klasifikasi yang cukup baik terhadap metadata artikel, serta berpotensi mendukung pengembangan sistem rekomendasi target submission jurnal The rapid growth of scientific publications in Indonesia has created the need for efficient methods to assist researchers in classifying and determining suitable journals for academic articles. Previous studies have shown that machine learning methods, particularly Naïve Bayes, are effective for various Indonesian text classification tasks. However, research specifically utilizing article metadata to determine the suitability of articles for SINTA-indexed journals remains limited, especially regarding the integration of TF–IDF features and cross-validation–based evaluation. This study aims to develop a classification model for determining article–journal suitability within SINTA using Term Frequency–Inverse Document Frequency and the Naïve Bayes algorithm. The dataset consists of 1,200 article metadata entries, including titles, abstracts, and keywords, collected through manual crawling of technology-related journals listed on the SINTA portal. The research stages include data collection, text preprocessing (case folding, translation, tokenization, stopword removal, and stemming), metadata merging, feature extraction using TF–IDF, and the implementation of Naïve Bayes with a 5-fold cross-validation scheme. Evaluation using confusion matrix metrics shows that the model achieved an accuracy of 0.7058, precision of 0.6977, recall of 0.7133, and an F1-score of 0.7065. These results indicate that Naïve Bayes provides a reasonably strong classification performance on article metadata and has potential application in journal submission recommendation systems

Keywords

• Artikel Akademik
• Machine Learning
• Publikasi Ilmiah
• Rekomendasi Jurnal
• Text Mining