Sistem Rekomendasi Penulis Bersama Berbasis Kedekatan Topik dan Riwayat Kolaborasi pada Metadata Publikasi

Budiman, Mohammad Idris Arif (2026) Sistem Rekomendasi Penulis Bersama Berbasis Kedekatan Topik dan Riwayat Kolaborasi pada Metadata Publikasi. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025221114-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Kolaborasi ilmiah penting untuk mempertemukan keahlian, metode, dan sudut pandang, tetapi pencarian kandidat penulis bersama pada institusi besar sering terhambat oleh tersebarnya informasi riset dalam metadata publikasi. Penelitian ini mengembangkan sistem rekomendasi penulis bersama berbasis kedekatan topik dan riwayat kolaborasi dengan penjelasan berbasis konten dan riwayat kolaborasi. Profil publikasi peneliti dibentuk dari judul, abstrak, dan kata kunci, direpresentasikan sebagai vektor semantik, lalu digunakan untuk memeringkatkan kandidat melalui kombinasi skor kedekatan topik dan skor riwayat kolaborasi berbobot waktu. Implementasi memakai metadata publikasi berafiliasi Institut Teknologi Sepuluh Nopember dengan pemisahan data pada tahun 2023, menghasilkan 10.968 publikasi data latih, 6.284 publikasi data uji, dan 2.860 peneliti pada ruang penulis data latih. Evaluasi dilakukan melalui pemulihan kolaborasi historis dan evaluasi periode uji. Pada evaluasi periode uji, metode yang menggabungkan kedekatan topik dan riwayat kolaborasi mencapai HR@10 sebesar 0,922, Recall@10 sebesar 0,541, NDCG@10 sebesar 0,580, dan MRR sebesar 0,740, lebih tinggi daripada metode berbasis konten, subprofil publikasi, dan popularitas. Penjelasan rekomendasi menyajikan bukti konten dan bukti riwayat kolaborasi agar alasan pemeringkatan dapat ditelusuri. Hasil ini terbatas pada evaluasi tanpa pelibatan pengguna langsung dan pada kandidat yang memiliki profil dalam data latih.
=======================================================================================================================================
Scientific collaboration is important for bringing together expertise, methods, and perspectives, yet identifying potential co-authors in large institutions is often constrained by dispersed research information in publication metadata. This study develops a co-author recommendation system based on topic proximity and collaboration history with explanations based on content and collaboration history. Researcher publication profiles are built from titles, abstracts, and keywords, represented as semantic embeddings, and used to rank candidates by combining topic proximity scores with temporally weighted collaboration history scores. The implementation uses publication metadata affiliated with Institut Teknologi Sepuluh Nopember with data split at year 2023, resulting in 10,968 training publications, 6,284 test publications, and 2,860 researchers in the training author pool. Evaluation is conducted through historical collaboration recovery and test period evaluation. In the test period evaluation, the method combining topic proximity and collaboration history achieves HR@10 of 0.922, Recall@10 of 0.541, NDCG@10 of 0.580, and MRR of 0.740, outperforming content-based and popularity-based methods, as well as a method based on publication subprofiles. The recommendation explanations present content evidence and collaboration-history evidence to ensure transparent ranking rationale. These results are limited to evaluation without direct user involvement and to candidates with profiles in the training data.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Kedekatan Topik, Rekomendasi Berbasis Penjelasan, Rekomendasi Penulis Bersama, Riwayat Kolaborasi, Co-Author Recommendation, Collaboration History, Explainable Recommendation, Topic Similarity
Subjects:	T Technology > T Technology (General) > T58.62 Decision support systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Mohammad Idris Arif Budiman
Date Deposited:	20 Jul 2026 07:58
Last Modified:	20 Jul 2026 07:58
URI:	http://repository.its.ac.id/id/eprint/135779

Actions (login required)

View Item