Faturrohman, Alif (2026) Adaptasi Domain Large Language Model Melalui Pendekatan RAFT Sebagai Layanan Backend Chatbot Edukatif Untuk Mendukung Literasi Ekonomi Syariah. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5026221040-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5026221040-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy |
Abstract
Tingkat literasi ekonomi syariah di Indonesia masih rendah, tercatat hanya 39,11% pada tahun 2024, menunjukkan adanya kesenjangan signifikan antara potensi industri halal dan pemahaman masyarakat. Tugas akhir ini bertujuan untuk mengembangkan chatbot edukatif berbasis Large Language Model (LLM) dengan pendekatan Retrieval-Augmented Fine-Tuning (RAFT) guna meningkatkan literasi ekonomi syariah remaja, serta membandingkan efektivitasnya dengan metode pembelajaran lain. Penelitian menggunakan metodologi Research and Development (R&D) dengan kerangka ADDIE. Evaluasi edukasi dilakukan pada 50 siswa yang terbagi dalam tiga kelompok: Modul cetak (n=18), LMS tanpa Chatbot (n=16), dan LMS + Chatbot (n=16). Peningkatan pemahaman diukur menggunakan N-Gain pada 10 soal yang identik antara pre-test dan post-test (auto-scoring). Hasil menunjukkan bahwa kelompok Modul tidak mengalami perubahan signifikan (N-Gain = -0,024, kategori Rendah), sementara kelompok LMS tanpa Chatbot (N-Gain = 0,646, Sedang dan LMS + Chatbot (N-Gain = 0,584, Sedang) menunjukkan peningkatan yang signifikan secara statistik dibandingkan Modul (p < 0,01). Namun, perbedaan antara LMS dengan dan tanpa chatbot tidak signifikan (p = 0,92), mengindikasikan bahwa efektivitas peningkatan pemahaman lebih disebabkan oleh penggunaan media pembelajaran digital secara umum dibandingkan fitur chatbot secara spesifik. Analisis berdasarkan perangkat (web vs mobile) tidak menunjukkan pola pengaruh yang sistematis. Validasi konten memperoleh rata-rata 4,28/5 (Sangat Baik) dan uji kepraktisan SUS sebesar 63,46/100 (Sedang).
=====================================================================================================================================
The level of sharia economic literacy in Indonesia remains low at only 39.11% in 2024, indicating a significant gap between the halal industry potential and public understanding. This final project aims to develop an educational chatbot based on Large Language Model (LLM) using Retrieval-Augmented Fine-Tuning (RAFT) approach to improve sharia economic literacy among adolescents, while comparing its effectiveness with other learning methods. The research employed Research and Development (R&D) methodology with the ADDIE framework. Educational evaluation was conducted on 50 students divided into three groups: printed module (n=18), LMS without Chatbot (n=16), and LMS + Chatbot (n=16). Learning improvement was measured using N-Gain on 10 identical questions between pre-test and post-test (auto-scoring). Results showed that the Module group had no significant change (N-Gain = -0.024, Low category), while both LMS without Chatbot (N-Gain = 0.646, Medium) and LMS + Chatbot (N-Gain = 0.584, Medium) demonstrated significant improvement compared to Module (p < 0.01). However, the difference between LMS with and without chatbot was not significant (p = 0.92), indicating that the effectiveness is attributed to digital learning media in general rather than the chatbot feature specifically. Device-based analysis (web vs mobile) showed no systematic influence pattern. Content validation achieved an average of 4.28/5 (Very Good) and SUS practicality score of 63,46/100 (Medium).
Actions (login required)
 |
View Item |