Pengembangan Sistem Informasi Repositori Penelitian Teknik Biomedik : Topic modeling dan Generative AI untuk Rekomendasi Riset

Hanifan, Anggara Izzat (2026) Pengembangan Sistem Informasi Repositori Penelitian Teknik Biomedik : Topic modeling dan Generative AI untuk Rekomendasi Riset. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5023221050_Undergraduate Thesis.pdf Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Perkembangan teknologi kecerdasan buatan membuka peluang baru dalam pengelolaan dan analisis dokumen ilmiah. Repositori institusi pada Departemen Teknik Biomedik ITS yang ada saat ini masih bersifat pasif dan hanya mengandalkan pencarian berbasis kata kunci tanpa pemahaman konteks semantik. Penelitian ini mengembangkan platform web BMEarchive yang mengintegrasikan dua pendekatan kecerdasan buatan, yaitu Domain Knowledge Graph-Based Topic modeling dengan Hybrid Similarity Engine untuk pemetaan topik, serta Retrieval-Augmented Generation (RAG) untuk asisten percakapan cerdas berbasis konteks repositori. Topic modeling diterapkan melalui pendekatan Domain Knowledge Graph-Based yang tidak hanya mengelompokkan dokumen, tetapi juga menghasilkan keyword fingerprint unik setiap topik menggunakan c-TF-IDF, membangun network graph interaktif yang memvisualisasikan hubungan antar penelitian, mengidentifikasi dokumen frontier sebagai indikator topik yang belum banyak dieksplorasi, serta memberikan rekomendasi penelitian serupa berbasis bobot kemiripan. Pendekatan ini menggabungkan kesamaan semantik berbasis Sentence-BERT dan kesamaan leksikal berbasis Kamus Teknik Biomedik melalui Hybrid Similarity Engine untuk memastikan pengelompokan yang akurat dan sesuai dengan terminologi teknis. Evaluasi kuantitatif topic modeling menunjukkan Topic Diversity 95.5%, NPMI Coherence 0.171, dan Graph Modularity 0.551. Evaluasi sistem RAG menggunakan kerangka RAGAS menghasilkan Faithfulness 0.740 dan Answer relevancy 0.840. Pengujian User Acceptance Testing (UAT) dengan 12 responden menghasilkan skor System Usability Scale (SUS) sebesar 81.25 (kategori Excellent, Grade A) serta rata-rata kepuasan fitur AI sebesar 4.37 dari skala 5.
================================================================================================================================
The advancement of artificial intelligence technology presents new opportunities in the management and analysis of scientific documents. The current institutional repository at the Department of Biomedical Engineering ITS remains passive and relies solely on keyword-based search without semantic context understanding. This research develops the BMEarchive web platform that integrates two artificial intelligence approaches: Domain Knowledge Graph-Based Topic modeling with Hybrid Similarity Engine for topic mapping, and Retrieval-Augmented Generation (RAG) for a context-aware intelligent conversational assistant based on repository content. The topic modeling is implemented through a Domain Knowledge Graph-Based approach that not only clusters documents but also generates unique keyword fingerprints for each topic using c-TF-IDF, builds an interactive network graph that visualizes relationships among research documents, identifies frontier documents as indicators of topics that have yet to be extensively explored, and provides similar research recommendations based on similarity weights. This approach combines semantic similarity based on Sentence-BERT and lexical similarity based on the Biomedical Engineering Dictionary through a Hybrid Similarity Engine to ensure accurate clustering aligned with domain-specific terminology. Quantitative evaluation of the topic modeling demonstrates Topic Diversity of 95.5%, NPMI Coherence of 0.171, and Graph Modularity of 0.551. Evaluation of the RAG system using the RAGAS framework yields Faithfulness of 0.740 and Answer relevancy of 0.840. User Acceptance Testing (UAT) involving 12 respondents produces a System Usability Scale (SUS) score of 81.25 (Excellent category, Grade A), along with an average AI feature satisfaction score of 4.37 out of 5.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Repositori ilmiah, Natural Language Processing, RAG, Topic modeling, Klastering, Keyword Extraction. Scientific repository, Natural Language Processing, RAG, Topic modeling, Cluster, Keyword Extraction.
Subjects:	T Technology > T Technology (General) T Technology > T Technology (General) > T385 Visualization--Technique T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T58.5 Information technology. IT--Auditing T Technology > T Technology (General) > T58.6 Management information systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Biomedical Engineering > 11410-(S1) Undergraduate Thesis
Depositing User:	Anggara Izzat Hanifan
Date Deposited:	01 Aug 2026 05:27
Last Modified:	01 Aug 2026 05:27
URI:	http://repository.its.ac.id/id/eprint/138698

Actions (login required)

View Item