Dashboard Cerdas Berbasis Large Language Model (LLM) Untuk Sistem Manajemen Aset

Ghifari, Miftah (2026) Dashboard Cerdas Berbasis Large Language Model (LLM) Untuk Sistem Manajemen Aset. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5024221019-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Pengelolaan aset Teknologi Informasi (IT) merupakan instrumen strategis yang krusial bagi keberlangsungan operasional perusahaan. PT Pertamina Trans Kontinental telah mendigitalisasi proses tersebut melalui sistem SAIL (System for Asset Information & Logistics), namun aksesibilitas informasi masih terhambat oleh navigasi antarmuka web yang kompleks bagi pengguna akhir.Penelitian ini bertujuan untuk merancang dan membangun sebuah dashboard cerdas berbasis Large Language Model (LLM) sebagai antarmuka percakapan intuitif guna meningkatkan efisiensi akses data aset. Sistem dikembangkan menggunakan arsitektur hibrida yang diorkestrasi oleh platform n8n, di mana logika navigasi diproses menggunakan model Llama 3 melalui API Groq untuk mencapai latensi rendah, sementara sintesis data sensitif dilakukan secara on-premise menggunakan model Qwen 2.5 melalui framework Ollama demi menjamin kedaulatan data.Metode Retrieval-Augmented Generation (RAG) diintegrasikan dengan basis data Microsoft SQL Server untuk memastikan respons bersifat faktual dan memitigasi risiko halusinasi informasi. Selain itu, sistem dioptimasi dengan teknik Robust Entity Extraction untuk menangani variasi input nonformal serta kesalahan pengetikan (typo), serta dilengkapi fitur Temporal Session Management untuk pengamanan token JWT pengguna.Hasil pengujian end-to-end menunjukkan bahwa sistem berhasil melakukan kueri informasi aset dengan tingkat akurasi mencapai 85% pada varian model Qwen 2.5-7B. Implementasi ini berhasil memangkas waktu pencarian informasi secara signifikan dibandingkan metode konvensional, sekaligus menyediakan fitur pelaporan dokumen PDF otomatis. Penelitian ini membuktikan bahwa integrasi LLM lokal dengan sistem manajemen aset korporasi mampu menciptakan layanan internal IT yang lebih responsif, aman, dan efisien.
=====================================================================================================================================
Information Technology (IT) asset management is a crucial strategic instrument for a company’s operational continuity. PT Pertamina Trans Kontinental has digitalized this process through the SAIL (System for Asset Information & Logistics) system, yet information accessibility remains hindered by complex web interface navigation for end-users. This research aims to design and develop a smart dashboard based on a Large Language Model (LLM) as an intuitive conversational interface to enhance asset data access efficiency. The system was developed using a hybrid architecture orchestrated by the n8n platform, where navigation logic is processed using the Llama 3 model via the Groq API to achieve low latency, while the synthesis of sensitive data is performed on-premise using the Qwen 2.5 model via the Ollama framework to ensure data sovereignty. The Retrieval-Augmented Generation (RAG) method is integrated with a Microsoft SQL Server database to ensure factual responses and mitigate the risk of information hallucination. Furthermore, the system is optimized with Robust Entity Extraction techniques to handle non-formal input variations and typos, and features Temporal Session Management for securing user JWT tokens. End-to-end testing results indicate that the system successfully performed asset information queries with an accuracy rate of up to 85% using the Qwen 2.5-7B model variant. This implementation significantly reduced information retrieval time compared to conventional methods and provided automated PDF document reporting features. This study demonstrates that integrating local LLMs with corporate asset management systems can create more responsive, secure, and efficient internal IT services.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Kata Kunci: Dashboard, Large Language Model (LLM), Retrieval-Augmented Generation (RAG), Hybrid AI, On-premise, SAIL ===================== Keywords: Dashboard, Large Language Model (LLM), Retrieval-Augmented Generation(RAG), Hybrid AI, On-premise, SAIL
Subjects:	Q Science > QA Mathematics > QA76.76.A65 Application software. Enterprise application integration (Computer systems)
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Computer Engineering > 90243-(S1) Undergraduate Thesis
Depositing User:	Miftah Ghifari
Date Deposited:	31 Jul 2026 03:59
Last Modified:	31 Jul 2026 03:59
URI:	http://repository.its.ac.id/id/eprint/140442

Actions (login required)

View Item