SISTEM REKOMENDASI REVIEWER ARTIKEL MENGGUNAKAN EUCLIDEAN DISTANCE SIMILARITY DENGAN METODE CONTENT-BASED COLLABORATIVE FILTERING  (Studi Kasus : ICVEE)

Article Reviewer Recommendation System Using Euclidean Distance Similarity with Content-Based Collaborative Filtering (Case Study: ICVEE)

Peningkatan jumlah penelitian dalam lingkungan akademik telah menghasilkan data yang kaya dan bernilai, khususnya dalam bentuk judul dan konten artikel ilmiah. Namun, sebagian besar institusi pendidikan masih mengelola data tersebut secara manual dan administratif, tanpa analisis lebih lanjut untuk mendukung pengambilan keputusan akademik. Dalam konteks ini, sistem rekomendasi reviewer berbasis content-based filtering menjadi solusi potensial untuk memetakan artikel dengan reviewer yang relevan berdasarkan kesamaan konten. Penelitian ini membangun sistem rekomendasi dengan pendekatan content-based menggunakan representasi teks TF-IDF dan pengukuran kemiripan Euclidean Distance. Setiap reviewer direpresentasikan melalui profil vektor berdasarkan artikel yang pernah direview, sementara artikel baru diubah menjadi vektor dan dibandingkan menggunakan kedua metode tersebut. Evaluasi dilakukan secara manual menggunakan 20 artikel sebagai data ground truth. Hasil evaluasi menunjukkan bahwa metode Euclidean Distance menghasilkan kinerja yang lebih baik dibandingkan Cosine Similarity, dengan akurasi sebesar 55%, precision 0.2333, recall 0.2121, dan F1-score 0.222. Sementara itu, metode Cosine menghasilkan akurasi 30%, precision 0.1167, recall 0.1061, dan F1-score 0.111. Kata Kunci : Sistem rekomendasi, Content-based filtering, TF-IDF, Euclidean Distance.

The increasing volume of research within academic environments has produced rich and valuable data, especially in the form of titles and contents of scientific articles. However, most educational institutions still manage this data manually and administratively, without further analysis to support academic decision-making. In this context, a content-based filtering recommendation system serves as a potential solution to map articles with relevant reviewers based on content similarity. This study develops a recommendation system using a content-based approach with TFIDF text representation and similarity measurement using Euclidean Distance. Each reviewer is represented through a vector profile based on previously reviewed articles, while new articles are transformed into vectors and compared using the two methods. The evaluation was conducted manually using 20 articles as ground truth data. The results show that the Euclidean Distance method outperforms Cosine Similarity, achieving an accuracy of 55%, a precision of 0.2333, recall of 0.2121, and an F1-score of 0.222. Meanwhile, the Cosine method achieved an accuracy of 30%, precision of 0.1167, recall of 0.1061, and an F1-score of 0.111. Keywords:Recommendation system, Content-based filtering, TF IDF, Euclidean Distance.