<![CDATA[Advances in artificial intelligence have driven the development of recommendation systems in the tourism sector, which is characterized by diverse destinations. This condition often makes it difficult for tourists to select destinations that match their preferences. Based on literature studies, Content-Based Filtering (CBF) is widely used due to its efficiency; however, it has limitations in understanding contextual information. In contrast, the Retrieval Augmented Generation (RAG) approach has been developed to improve recommendation quality through semantic understanding. This study aims to compare the performance of CBF and RAG in tourism destination recommendation systems. CBF employs TF-IDF and cosine similarity to measure content similarity, while RAG integrates retrieval and generation processes using the LLaMA 3.2 model and the FAISS vector database. The research methodology includes data collection, text preprocessing, system implementation, and evaluation using context recall, faithfulness, answer relevancy, and similarity metrics. The results indicate that CBF achieved a context recall of 0.317, faithfulness of 1.000, answer relevancy of 0.190, and similarity of 0.293, demonstrating high accuracy with respect to source data but limited contextual understanding. Meanwhile, RAG achieved a context recall of 1.000, faithfulness of 0.783, answer relevancy of 0.617, and similarity of 0.715, indicating superior performance in generating relevant recommendations. In conclusion, RAG outperforms CBF in contextual and semantic aspects, while CBF remains more efficient in processing explicit data. This study is expected to serve as a reference for developing more adaptive and personalized tourism recommendation systems]]>
Copyrights © 2026
