Implementasi Large Language Models Gemini Pada Chatbot Penerimaan Mahasiswa Baru STMIK Widya Cipta Dharma Berbasis Website

Anwar, Faidil (2026) Implementasi Large Language Models Gemini Pada Chatbot Penerimaan Mahasiswa Baru STMIK Widya Cipta Dharma Berbasis Website. S1 Teknik Informatika thesis, STMIK Widya Cipta Dharma.

	Text 2243040-S1-Jurnal.pdf Download (1MB)
	Text 2243040-S1-Teknik Informatika.pdf Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Layanan Penerimaan Mahasiswa Baru (PMB) di STMIK Widya Cipta Dharma memerlukan penyampaian informasi yang komprehensif dan dapat diakses kapan saja oleh calon mahasiswa. Namun, pelayanan konvensional terkendala batas jam kerja operasional serta tingginya beban kerja staf untuk menjawab pertanyaan berulang. Meskipun Large Language Models (LLMs) mampu memberikan interaksi komunikatif, teknologi ini memiliki risiko halusinasi informasi. Untuk mengatasi permasalahan tersebut, penelitian ini bertujuan merancang dan mengimplementasikan chatbot PMB berbasis website menggunakan Gemini LLM yang diintegrasikan dengan arsitektur Retrieval-Augmented Generation (RAG) agar informasi yang disampaikan selalu akurat sesuai dokumen resmi kampus. Pengembangan sistem menerapkan metode Rapid Application Development (RAD) yang meliputi tahapan Requirements Planning, User Design, Construction, dan Cutover. Chatbot ini mengandalkan model gemini-3.1-flash-lite sebagai mesin pemrosesan bahasa alami utama, yang didukung oleh Qdrant Vector Database untuk pencarian semantik dan indeks konteks dokumen basis pengetahuan kampus. Hasil evaluasi menunjukkan bahwa sistem berhasil diimplementasikan secara optimal. Pengujian Black Box memvalidasi keberhasilan seluruh fitur fungsional masukan dan keluaran. Sementara itu, White Box Testing membuktikan alur eksekusi pesan bebas dari kesalahan dengan jalur kompleksitas siklomatis yang valid. Selain itu, evaluasi Beta Testing menghasilkan tingkat kepuasan pengguna sebesar 88,8% dari responden calon mahasiswa. Kesimpulannya, implementasi Gemini LLM dan RAG pada chatbot PMB ini terbukti layak, akurat, serta efektif dalam memberikan layanan informasi 24 jam tanpa risiko halusinasi. ====================================================================================================== The New Student Admission (PMB) service at STMIK Widya Cipta Dharma requires comprehensive and continuously accessible information delivery for prospective students. However, conventional services face operational working hour limitations and high staff workloads from repetitive inquiries. While Large Language Models (LLMs) offer interactive communication, they present hallucination risks. To address these challenges, this study designs and implements a website-based PMB chatbot using the Gemini LLM integrated with a Retrieval-Augmented Generation (RAG) architecture to guarantee factual accuracy grounded in official campus documents. Using the Rapid Application Development (RAD) methodology, development progressed through Requirements Planning, User Design, Construction, and Cutover phases. The system employs the gemini-3.1-flash-lite model for natural language processing, complemented by the Qdrant Vector Database to handle semantic search and context indexing for the campus knowledge base. System evaluation confirmed successful implementation across multiple testing phases. Black Box Testing verified that all functional features perform as expected, while White Box Testing confirmed an error-free execution flow with valid cyclomatic complexity paths. Additionally, Beta Testing demonstrated high user acceptance, achieving an 88.8% satisfaction rate among prospective student respondents. Ultimately, combining the Gemini LLM with RAG architecture proves to be a feasible, highly accurate, and effective solution for delivering 24-hour admissions information without the risk of AI hallucination.

Item Type:	Thesis (S1 Teknik Informatika)
Additional Information:	Pembimbing 1 : Ita Arfyanti, S.Kom., M.M Pembimbing 2 : Ahmad Fahrijal Pukeng, S.Kom., M.T
Uncontrolled Keywords:	Large Language Models, Gemini, Chatbot, Retrieval-Augmented Generation, Website
Subjects:	Q Science > QA Mathematics > QA76 Computer software
Divisions:	Teknik Informatika
Depositing User:	Mr Faidil Anwar
Date Deposited:	03 Aug 2026 02:52
Last Modified:	03 Aug 2026 02:52
URI:	http://repository.wicida.ac.id/id/eprint/6332

Actions (login required)

View Item