Implementasi Retrieval-Augmented Generation (RAG) pada Sistem Rekomendasi Restoran Berdasarkan Ulasan

Ulinnuha, Hasna Daffa (2026) Implementasi Retrieval-Augmented Generation (RAG) pada Sistem Rekomendasi Restoran Berdasarkan Ulasan. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5026221205-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Ulasan Google Maps mengandung informasi mengenai pengalaman pelanggan terhadap berbagai aspek restoran, seperti food, service, price, dan ambience. Namun, sebagian besar sistem rekomendasi restoran masih mengandalkan data terstruktur, sehingga informasi yang terkandung dalam ulasan belum dimanfaatkan secara optimal. Penelitian ini bertujuan mengembangkan sistem rekomendasi restoran berbasis Retrieval-Augmented Generation (RAG) yang memanfaatkan hasil Aspect-Based Sentiment Analysis (ABSA) untuk menghasilkan rekomendasi yang lebih kontekstual sesuai preferensi pengguna. Penelitian diawali dengan pengumpulan metadata dan ulasan dari 125 restoran di Surabaya serta anotasi terhadap 1.800 ulasan. Selanjutnya dikembangkan model Aspect Detection dan Sentiment Classification, kemudian hasil prediksi aspek dan sentimen diintegrasikan bersama metadata restoran ke dalam knowledge base sistem RAG. Sistem yang dikembangkan mendukung rekomendasi berdasarkan metadata restoran maupun ulasan pelanggan.Hasil penelitian menunjukkan bahwa Linear Support Vector Machine (Linear SVM) memperoleh performa terbaik pada tugas Aspect Detection dengan Macro F1-score sebesar 0,8313, sedangkan IndoBERT memberikan performa terbaik pada tugas Sentiment Classification dengan Accuracy sebesar 80,07% dan Macro F1-score sebesar 72,83%. Evaluasi retrieval pada sistem RAG berbasis metadata menghasilkan nilai Hit@1, Hit@3, Hit@5, Hit@10, dan MRR sebesar 1,00, sedangkan evaluasi sistem berbasis ulasan memperoleh rata-rata Human Evaluation sebesar 4,30 dari skala 5. Hasil tersebut menunjukkan bahwa sistem mampu mengambil informasi yang relevan serta memberikan rekomendasi restoran yang sesuai dengan kebutuhan pengguna.
======================================================================================================================================
Google Maps reviews contain valuable information about customers' experiences with various restaurant aspects, including food, service, price, and ambience. However, most restaurant recommendation systems still rely on structured data, such as ratings and restaurant categories, limiting the utilization of information embedded in textual reviews. This study aims to develop a Retrieval-Augmented Generation (RAG)-based restaurant recommendation system that leverages Aspect-Based Sentiment Analysis (ABSA) to generate more contextual recommendations aligned with users' preferences. The study began by collecting metadata and reviews from 125 restaurants in Surabaya, followed by the annotation of 1,800 reviews. Subsequently, Aspect Detection and Sentiment Classification models were developed, and the predicted aspect and sentiment labels were integrated with restaurant metadata into the RAG knowledge base. The proposed system supports two recommendation scenarios: recommendations based on restaurant metadata and recommendations based on customer reviews. The experimental results show that Linear Support Vector Machine (Linear SVM) achieved the best performance for the Aspect Detection task, with a Macro F1-score of 0.8313. Meanwhile, IndoBERT achieved the best performance for the Sentiment Classification task, obtaining an Accuracy of 80.07% and a Macro F1-score of 72.83%. The metadata-based RAG retrieval achieved Hit@1, Hit@3, Hit@5, Hit@10, and Mean Reciprocal Rank (MRR) scores of 1.00, while the review-based recommendation system obtained an average Human Evaluation score of 4.30 out of 5. These results demonstrate that the proposed system is capable of retrieving relevant information and providing restaurant recommendations that effectively meet users' preferences.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Aspect-Based Sentiment Analysis, Retrieval-Augmented Generation, Sistem Rekomendasi Restoran, IndoBERT Retrieval-Augmented Generation (RAG), Aspect-Based Sentiment Analysis (ABSA), Restaurant Recommendation System, IndoBERT.
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis
Depositing User:	Hasna Daffa Ulinnuha
Date Deposited:	01 Aug 2026 04:29
Last Modified:	01 Aug 2026 04:29
URI:	http://repository.its.ac.id/id/eprint/141284

Actions (login required)

View Item